Downloads: 1
2084 lines (2083 with data), 174.0 kB
{
"cells": [
{
"cell_type": "markdown",
"id": "b348fcda-1cd7-424f-aafb-f0fb5f4fe94d",
"metadata": {},
"source": [
"### Loading dataset"
]
},
{
"cell_type": "code",
"execution_count": 1,
"id": "564c5a25-bc61-4d3d-acca-d54d998bdb4a",
"metadata": {},
"outputs": [
{
"data": {
"text/html": [
"<div>\n",
"<style scoped>\n",
" .dataframe tbody tr th:only-of-type {\n",
" vertical-align: middle;\n",
" }\n",
"\n",
" .dataframe tbody tr th {\n",
" vertical-align: top;\n",
" }\n",
"\n",
" .dataframe thead th {\n",
" text-align: right;\n",
" }\n",
"</style>\n",
"<table border=\"1\" class=\"dataframe\">\n",
" <thead>\n",
" <tr style=\"text-align: right;\">\n",
" <th></th>\n",
" <th>ID</th>\n",
" <th>pCR (outcome)</th>\n",
" <th>RelapseFreeSurvival (outcome)</th>\n",
" <th>Age</th>\n",
" <th>ER</th>\n",
" <th>PgR</th>\n",
" <th>HER2</th>\n",
" <th>TrippleNegative</th>\n",
" <th>ChemoGrade</th>\n",
" <th>Proliferation</th>\n",
" <th>...</th>\n",
" <th>original_glszm_SmallAreaHighGrayLevelEmphasis</th>\n",
" <th>original_glszm_SmallAreaLowGrayLevelEmphasis</th>\n",
" <th>original_glszm_ZoneEntropy</th>\n",
" <th>original_glszm_ZonePercentage</th>\n",
" <th>original_glszm_ZoneVariance</th>\n",
" <th>original_ngtdm_Busyness</th>\n",
" <th>original_ngtdm_Coarseness</th>\n",
" <th>original_ngtdm_Complexity</th>\n",
" <th>original_ngtdm_Contrast</th>\n",
" <th>original_ngtdm_Strength</th>\n",
" </tr>\n",
" </thead>\n",
" <tbody>\n",
" <tr>\n",
" <th>0</th>\n",
" <td>TRG002174</td>\n",
" <td>1</td>\n",
" <td>144.0</td>\n",
" <td>41.0</td>\n",
" <td>0</td>\n",
" <td>0</td>\n",
" <td>0</td>\n",
" <td>1</td>\n",
" <td>3</td>\n",
" <td>3</td>\n",
" <td>...</td>\n",
" <td>0.517172</td>\n",
" <td>0.375126</td>\n",
" <td>3.325332</td>\n",
" <td>0.002314</td>\n",
" <td>3880771.500</td>\n",
" <td>473.464852</td>\n",
" <td>0.000768</td>\n",
" <td>0.182615</td>\n",
" <td>0.030508</td>\n",
" <td>0.000758</td>\n",
" </tr>\n",
" <tr>\n",
" <th>1</th>\n",
" <td>TRG002178</td>\n",
" <td>0</td>\n",
" <td>142.0</td>\n",
" <td>39.0</td>\n",
" <td>1</td>\n",
" <td>1</td>\n",
" <td>0</td>\n",
" <td>0</td>\n",
" <td>3</td>\n",
" <td>3</td>\n",
" <td>...</td>\n",
" <td>0.444391</td>\n",
" <td>0.444391</td>\n",
" <td>3.032144</td>\n",
" <td>0.005612</td>\n",
" <td>2372009.744</td>\n",
" <td>59.459710</td>\n",
" <td>0.004383</td>\n",
" <td>0.032012</td>\n",
" <td>0.001006</td>\n",
" <td>0.003685</td>\n",
" </tr>\n",
" <tr>\n",
" <th>2</th>\n",
" <td>TRG002204</td>\n",
" <td>1</td>\n",
" <td>135.0</td>\n",
" <td>31.0</td>\n",
" <td>0</td>\n",
" <td>0</td>\n",
" <td>0</td>\n",
" <td>1</td>\n",
" <td>2</td>\n",
" <td>1</td>\n",
" <td>...</td>\n",
" <td>0.534549</td>\n",
" <td>0.534549</td>\n",
" <td>2.485848</td>\n",
" <td>0.006752</td>\n",
" <td>1540027.421</td>\n",
" <td>33.935384</td>\n",
" <td>0.007584</td>\n",
" <td>0.024062</td>\n",
" <td>0.000529</td>\n",
" <td>0.006447</td>\n",
" </tr>\n",
" <tr>\n",
" <th>3</th>\n",
" <td>TRG002206</td>\n",
" <td>0</td>\n",
" <td>12.0</td>\n",
" <td>35.0</td>\n",
" <td>0</td>\n",
" <td>0</td>\n",
" <td>0</td>\n",
" <td>1</td>\n",
" <td>3</td>\n",
" <td>3</td>\n",
" <td>...</td>\n",
" <td>0.506185</td>\n",
" <td>0.506185</td>\n",
" <td>2.606255</td>\n",
" <td>0.003755</td>\n",
" <td>6936740.794</td>\n",
" <td>46.859265</td>\n",
" <td>0.005424</td>\n",
" <td>0.013707</td>\n",
" <td>0.000178</td>\n",
" <td>0.004543</td>\n",
" </tr>\n",
" <tr>\n",
" <th>4</th>\n",
" <td>TRG002210</td>\n",
" <td>0</td>\n",
" <td>109.0</td>\n",
" <td>61.0</td>\n",
" <td>1</td>\n",
" <td>0</td>\n",
" <td>0</td>\n",
" <td>0</td>\n",
" <td>2</td>\n",
" <td>1</td>\n",
" <td>...</td>\n",
" <td>0.462282</td>\n",
" <td>0.462282</td>\n",
" <td>2.809279</td>\n",
" <td>0.006521</td>\n",
" <td>1265399.054</td>\n",
" <td>39.621023</td>\n",
" <td>0.006585</td>\n",
" <td>0.034148</td>\n",
" <td>0.001083</td>\n",
" <td>0.005626</td>\n",
" </tr>\n",
" </tbody>\n",
"</table>\n",
"<p>5 rows × 121 columns</p>\n",
"</div>"
],
"text/plain": [
" ID pCR (outcome) RelapseFreeSurvival (outcome) Age ER PgR \\\n",
"0 TRG002174 1 144.0 41.0 0 0 \n",
"1 TRG002178 0 142.0 39.0 1 1 \n",
"2 TRG002204 1 135.0 31.0 0 0 \n",
"3 TRG002206 0 12.0 35.0 0 0 \n",
"4 TRG002210 0 109.0 61.0 1 0 \n",
"\n",
" HER2 TrippleNegative ChemoGrade Proliferation ... \\\n",
"0 0 1 3 3 ... \n",
"1 0 0 3 3 ... \n",
"2 0 1 2 1 ... \n",
"3 0 1 3 3 ... \n",
"4 0 0 2 1 ... \n",
"\n",
" original_glszm_SmallAreaHighGrayLevelEmphasis \\\n",
"0 0.517172 \n",
"1 0.444391 \n",
"2 0.534549 \n",
"3 0.506185 \n",
"4 0.462282 \n",
"\n",
" original_glszm_SmallAreaLowGrayLevelEmphasis original_glszm_ZoneEntropy \\\n",
"0 0.375126 3.325332 \n",
"1 0.444391 3.032144 \n",
"2 0.534549 2.485848 \n",
"3 0.506185 2.606255 \n",
"4 0.462282 2.809279 \n",
"\n",
" original_glszm_ZonePercentage original_glszm_ZoneVariance \\\n",
"0 0.002314 3880771.500 \n",
"1 0.005612 2372009.744 \n",
"2 0.006752 1540027.421 \n",
"3 0.003755 6936740.794 \n",
"4 0.006521 1265399.054 \n",
"\n",
" original_ngtdm_Busyness original_ngtdm_Coarseness \\\n",
"0 473.464852 0.000768 \n",
"1 59.459710 0.004383 \n",
"2 33.935384 0.007584 \n",
"3 46.859265 0.005424 \n",
"4 39.621023 0.006585 \n",
"\n",
" original_ngtdm_Complexity original_ngtdm_Contrast original_ngtdm_Strength \n",
"0 0.182615 0.030508 0.000758 \n",
"1 0.032012 0.001006 0.003685 \n",
"2 0.024062 0.000529 0.006447 \n",
"3 0.013707 0.000178 0.004543 \n",
"4 0.034148 0.001083 0.005626 \n",
"\n",
"[5 rows x 121 columns]"
]
},
"execution_count": 1,
"metadata": {},
"output_type": "execute_result"
}
],
"source": [
"import pandas as pd\n",
"\n",
"df= pd.read_excel(\"TrainDataset2024.xls\")\n",
"df.head()"
]
},
{
"cell_type": "markdown",
"id": "6cecd481-9f9e-45a6-9ece-6b43a71b2fb1",
"metadata": {},
"source": [
"## Data Preprocessing"
]
},
{
"cell_type": "code",
"execution_count": 4,
"id": "52341c70-e6d4-4738-8817-eda5398912d8",
"metadata": {},
"outputs": [],
"source": [
"df = df.rename(columns={\"pCR (outcome)\": \"PCR\"})"
]
},
{
"cell_type": "code",
"execution_count": 5,
"id": "8457af35-f78b-4595-b162-77fa6f62621d",
"metadata": {},
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
"Number of null values: 105\n"
]
}
],
"source": [
"import numpy as np\n",
"df.replace(999, np.nan, inplace= True)\n",
"null = df.isna().sum().sum()\n",
"print(\"Number of null values: \",null)"
]
},
{
"cell_type": "code",
"execution_count": 6,
"id": "cf17629c-72ca-4fb7-a094-8c26cfd34dfa",
"metadata": {},
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
"\n",
"=== Features with Missing Values ===\n",
"PCR 5\n",
"PgR 1\n",
"HER2 1\n",
"TrippleNegative 1\n",
"ChemoGrade 3\n",
"Proliferation 2\n",
"HistologyType 3\n",
"LNStatus 1\n",
"Gene 88\n",
"dtype: int64\n"
]
},
{
"data": {
"image/png": "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",
"text/plain": [
"<Figure size 1200x600 with 1 Axes>"
]
},
"metadata": {},
"output_type": "display_data"
},
{
"name": "stdout",
"output_type": "stream",
"text": [
"\n",
"=== Overall Missing Data Summary ===\n",
"Total number of missing values: 105\n",
"Total number of cells in dataset: 48400\n",
"Percentage of missing data: 0.22%\n"
]
}
],
"source": [
"import numpy as np\n",
"import pandas as pd\n",
"import matplotlib.pyplot as plt\n",
"\n",
"# Replace 999 with NaN for further analysis\n",
"df.replace(999, np.nan, inplace=True)\n",
"\n",
"# Count missing (999) values in each column\n",
"missing_counts = (df.isna().sum())\n",
"\n",
"# Filter out columns with no missing values\n",
"missing_counts = missing_counts[missing_counts > 0]\n",
"\n",
"# Display feature names and missing value counts\n",
"print(\"\\n=== Features with Missing Values ===\")\n",
"print(missing_counts)\n",
"\n",
"# Visualize the distribution of missing values\n",
"plt.figure(figsize=(12, 6))\n",
"missing_counts.plot(kind='bar', color='skyblue', edgecolor='black')\n",
"plt.title('Number of Missing Values by Feature')\n",
"plt.xlabel('Features')\n",
"plt.ylabel('Count of Missing Values')\n",
"plt.xticks(rotation=45, ha='right')\n",
"plt.tight_layout()\n",
"plt.show()\n",
"\n",
"# Summary statistics\n",
"total_missing = missing_counts.sum()\n",
"total_cells = df.size\n",
"missing_percentage = (total_missing / total_cells) * 100\n",
"\n",
"print(\"\\n=== Overall Missing Data Summary ===\")\n",
"print(f\"Total number of missing values: {total_missing}\")\n",
"print(f\"Total number of cells in dataset: {total_cells}\")\n",
"print(f\"Percentage of missing data: {missing_percentage:.2f}%\")\n"
]
},
{
"cell_type": "code",
"execution_count": 7,
"id": "68cedf35-6386-426c-ac64-1127dd4c9935",
"metadata": {},
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
"PCR 0\n",
"PgR 0\n",
"HER2 0\n",
"TrippleNegative 0\n",
"ChemoGrade 0\n",
"Proliferation 0\n",
"HistologyType 0\n",
"LNStatus 0\n",
"Gene 0\n",
"dtype: int64\n"
]
},
{
"name": "stderr",
"output_type": "stream",
"text": [
"C:\\Users\\LLR User\\AppData\\Local\\Temp\\ipykernel_14388\\4226577187.py:12: FutureWarning: A value is trying to be set on a copy of a DataFrame or Series through chained assignment using an inplace method.\n",
"The behavior will change in pandas 3.0. This inplace method will never work because the intermediate object on which we are setting values always behaves as a copy.\n",
"\n",
"For example, when doing 'df[col].method(value, inplace=True)', try using 'df.method({col: value}, inplace=True)' or df[col] = df[col].method(value) instead, to perform the operation inplace on the original object.\n",
"\n",
"\n",
" df[feature].fillna(mode_value, inplace=True) # Replace NaN with the mode\n"
]
}
],
"source": [
"import pandas as pd\n",
"\n",
"# List of features you want to modify\n",
"features_to_modify = [\n",
" \"PCR\", \"PgR\", \"HER2\", \"TrippleNegative\", \"ChemoGrade\", \n",
" \"Proliferation\", \"HistologyType\", \"LNStatus\", \"Gene\"\n",
"]\n",
"\n",
"# Loop through the columns and replace NaN values with the mode of each column\n",
"for feature in features_to_modify:\n",
" mode_value = df[feature].mode()[0] # Get the mode of the feature\n",
" df[feature].fillna(mode_value, inplace=True) # Replace NaN with the mode\n",
"\n",
"# Check the changes\n",
"print(df[features_to_modify].isna().sum()) # Check how many NaN values are left\n"
]
},
{
"cell_type": "code",
"execution_count": 8,
"id": "995b86b2-4475-4b94-8a7e-044bebaf8556",
"metadata": {},
"outputs": [],
"source": [
"df = df.drop('ID', axis=1)"
]
},
{
"cell_type": "code",
"execution_count": 9,
"id": "2f8b60cc-c014-4424-ba37-d7705812d8bc",
"metadata": {},
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
"Number of outliers: 87\n",
"Number of outliers: PCR 65.000000\n",
"RelapseFreeSurvival (outcome) 17185.083333\n",
"Age 16080.576595\n",
"ER 166.000000\n",
"PgR 126.000000\n",
" ... \n",
"original_ngtdm_Busyness 36656.435955\n",
"original_ngtdm_Coarseness 4.608712\n",
"original_ngtdm_Complexity 18.424557\n",
"original_ngtdm_Contrast 1.797244\n",
"original_ngtdm_Strength 4.089319\n",
"Length: 120, dtype: float64\n"
]
}
],
"source": [
"# Calculate Z-scores for the entire dataset\n",
"z_scores = np.abs((df - df.mean()) / df.std())\n",
"\n",
"# Identify rows that have z-scores above the threshold (3 in this case)\n",
"outliers = (z_scores > 3).any(axis=1)\n",
"\n",
"# Print the number of outliers\n",
"print(f\"Number of outliers: {outliers.sum()}\")\n",
"\n",
"# Remove outliers from the dataset\n",
"df_no_outliers = df[~outliers]\n",
"\n",
"print(f\"Number of outliers: {df_no_outliers.sum()}\")"
]
},
{
"cell_type": "markdown",
"id": "37f55653-9445-404e-8730-c1c93c45e00f",
"metadata": {},
"source": [
"### Feature Selection"
]
},
{
"cell_type": "code",
"execution_count": 10,
"id": "70899e82-33c6-495d-baaf-e8d6d5b4afa1",
"metadata": {},
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
"correlated features: 82\n"
]
}
],
"source": [
"df1 = df.copy()\n",
"# checking for correlated features of dataset\n",
"def correlation(data, threshold):\n",
" col_corr = {} # Dictionary to store correlated features\n",
" corr_matrix = data.corr()\n",
" for i in range(len(corr_matrix.columns)):\n",
" for j in range(i):\n",
" if abs(corr_matrix.iloc[i, j]) > threshold: # We are interested in absolute coefficient value\n",
" colname = corr_matrix.columns[i]\n",
" if colname not in col_corr:\n",
" col_corr[colname] = set()\n",
" col_corr[colname].add(corr_matrix.columns[j])\n",
"\n",
" return col_corr\n",
"\n",
"corr_features = correlation(df1, 0.8)\n",
"print('correlated features: ', len(corr_features))"
]
},
{
"cell_type": "code",
"execution_count": 11,
"id": "e8695d75-11cd-4f48-b301-d019ce147584",
"metadata": {},
"outputs": [
{
"data": {
"text/plain": [
"(400, 38)"
]
},
"execution_count": 11,
"metadata": {},
"output_type": "execute_result"
}
],
"source": [
"# removing the correlated features\n",
"df_corr= df1.drop(labels=corr_features, axis=1)\n",
"df_corr.shape"
]
},
{
"cell_type": "code",
"execution_count": 12,
"id": "164fcddb-b51c-41b7-b23a-a4f6c7725c84",
"metadata": {},
"outputs": [
{
"data": {
"image/png": "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",
"text/plain": [
"<Figure size 640x480 with 1 Axes>"
]
},
"metadata": {},
"output_type": "display_data"
}
],
"source": [
"# checking the count of target values after data processing and feature selection\n",
"outcomes = df_corr['PCR'].value_counts()\n",
"outcome_labels =['0','1']\n",
"outcome_values = outcomes.values\n",
"plt.bar(outcome_labels, outcome_values)\n",
"plt.xlabel('Outcome')\n",
"plt.ylabel('Count')\n",
"plt.title('Distribution of Outcomes in PCR')\n",
"plt.show()"
]
},
{
"cell_type": "code",
"execution_count": 13,
"id": "8f4b2fc4-664b-4c47-af59-e5646acb9aed",
"metadata": {},
"outputs": [],
"source": [
"# dataset using correlated feature selection\n",
"X_corr = df_corr.drop([\"PCR\",\"RelapseFreeSurvival (outcome)\"],axis=1)\n",
"y_corr = df_corr[\"PCR\"]"
]
},
{
"cell_type": "code",
"execution_count": 14,
"id": "683c242f-c3ba-4268-8318-eb01d4f1d9aa",
"metadata": {},
"outputs": [
{
"data": {
"text/html": [
"<div>\n",
"<style scoped>\n",
" .dataframe tbody tr th:only-of-type {\n",
" vertical-align: middle;\n",
" }\n",
"\n",
" .dataframe tbody tr th {\n",
" vertical-align: top;\n",
" }\n",
"\n",
" .dataframe thead th {\n",
" text-align: right;\n",
" }\n",
"</style>\n",
"<table border=\"1\" class=\"dataframe\">\n",
" <thead>\n",
" <tr style=\"text-align: right;\">\n",
" <th></th>\n",
" <th>Age</th>\n",
" <th>ER</th>\n",
" <th>PgR</th>\n",
" <th>HER2</th>\n",
" <th>TrippleNegative</th>\n",
" <th>ChemoGrade</th>\n",
" <th>HistologyType</th>\n",
" <th>LNStatus</th>\n",
" <th>TumourStage</th>\n",
" <th>Gene</th>\n",
" <th>...</th>\n",
" <th>original_gldm_SmallDependenceEmphasis</th>\n",
" <th>original_glrlm_LongRunLowGrayLevelEmphasis</th>\n",
" <th>original_glrlm_ShortRunHighGrayLevelEmphasis</th>\n",
" <th>original_glszm_GrayLevelNonUniformity</th>\n",
" <th>original_glszm_GrayLevelNonUniformityNormalized</th>\n",
" <th>original_glszm_LargeAreaEmphasis</th>\n",
" <th>original_glszm_SizeZoneNonUniformityNormalized</th>\n",
" <th>original_glszm_SmallAreaEmphasis</th>\n",
" <th>original_ngtdm_Busyness</th>\n",
" <th>original_ngtdm_Strength</th>\n",
" </tr>\n",
" </thead>\n",
" <tbody>\n",
" <tr>\n",
" <th>0</th>\n",
" <td>41.0</td>\n",
" <td>0</td>\n",
" <td>0.0</td>\n",
" <td>0.0</td>\n",
" <td>1.0</td>\n",
" <td>3.0</td>\n",
" <td>1.0</td>\n",
" <td>1.0</td>\n",
" <td>2</td>\n",
" <td>1.0</td>\n",
" <td>...</td>\n",
" <td>0.005563</td>\n",
" <td>10.779989</td>\n",
" <td>0.789987</td>\n",
" <td>27.545455</td>\n",
" <td>0.834711</td>\n",
" <td>4067578.818</td>\n",
" <td>0.180900</td>\n",
" <td>0.403535</td>\n",
" <td>473.464852</td>\n",
" <td>0.000758</td>\n",
" </tr>\n",
" <tr>\n",
" <th>1</th>\n",
" <td>39.0</td>\n",
" <td>1</td>\n",
" <td>1.0</td>\n",
" <td>0.0</td>\n",
" <td>0.0</td>\n",
" <td>3.0</td>\n",
" <td>1.0</td>\n",
" <td>1.0</td>\n",
" <td>2</td>\n",
" <td>0.0</td>\n",
" <td>...</td>\n",
" <td>0.006518</td>\n",
" <td>27.650685</td>\n",
" <td>0.442279</td>\n",
" <td>78.025000</td>\n",
" <td>0.975313</td>\n",
" <td>2403756.075</td>\n",
" <td>0.198125</td>\n",
" <td>0.444391</td>\n",
" <td>59.459710</td>\n",
" <td>0.003685</td>\n",
" </tr>\n",
" <tr>\n",
" <th>2</th>\n",
" <td>31.0</td>\n",
" <td>0</td>\n",
" <td>0.0</td>\n",
" <td>0.0</td>\n",
" <td>1.0</td>\n",
" <td>2.0</td>\n",
" <td>1.0</td>\n",
" <td>0.0</td>\n",
" <td>2</td>\n",
" <td>1.0</td>\n",
" <td>...</td>\n",
" <td>0.007181</td>\n",
" <td>25.338218</td>\n",
" <td>0.503046</td>\n",
" <td>72.027027</td>\n",
" <td>0.973338</td>\n",
" <td>1561963.432</td>\n",
" <td>0.275749</td>\n",
" <td>0.534549</td>\n",
" <td>33.935384</td>\n",
" <td>0.006447</td>\n",
" </tr>\n",
" <tr>\n",
" <th>3</th>\n",
" <td>35.0</td>\n",
" <td>0</td>\n",
" <td>0.0</td>\n",
" <td>0.0</td>\n",
" <td>1.0</td>\n",
" <td>3.0</td>\n",
" <td>1.0</td>\n",
" <td>1.0</td>\n",
" <td>3</td>\n",
" <td>1.0</td>\n",
" <td>...</td>\n",
" <td>0.004902</td>\n",
" <td>31.461354</td>\n",
" <td>0.399896</td>\n",
" <td>99.019802</td>\n",
" <td>0.980394</td>\n",
" <td>7007670.723</td>\n",
" <td>0.253014</td>\n",
" <td>0.506185</td>\n",
" <td>46.859265</td>\n",
" <td>0.004543</td>\n",
" </tr>\n",
" <tr>\n",
" <th>4</th>\n",
" <td>61.0</td>\n",
" <td>1</td>\n",
" <td>0.0</td>\n",
" <td>0.0</td>\n",
" <td>0.0</td>\n",
" <td>2.0</td>\n",
" <td>1.0</td>\n",
" <td>0.0</td>\n",
" <td>2</td>\n",
" <td>1.0</td>\n",
" <td>...</td>\n",
" <td>0.007222</td>\n",
" <td>27.916261</td>\n",
" <td>0.473278</td>\n",
" <td>56.034483</td>\n",
" <td>0.966112</td>\n",
" <td>1288913.690</td>\n",
" <td>0.216409</td>\n",
" <td>0.462282</td>\n",
" <td>39.621023</td>\n",
" <td>0.005626</td>\n",
" </tr>\n",
" </tbody>\n",
"</table>\n",
"<p>5 rows × 36 columns</p>\n",
"</div>"
],
"text/plain": [
" Age ER PgR HER2 TrippleNegative ChemoGrade HistologyType LNStatus \\\n",
"0 41.0 0 0.0 0.0 1.0 3.0 1.0 1.0 \n",
"1 39.0 1 1.0 0.0 0.0 3.0 1.0 1.0 \n",
"2 31.0 0 0.0 0.0 1.0 2.0 1.0 0.0 \n",
"3 35.0 0 0.0 0.0 1.0 3.0 1.0 1.0 \n",
"4 61.0 1 0.0 0.0 0.0 2.0 1.0 0.0 \n",
"\n",
" TumourStage Gene ... original_gldm_SmallDependenceEmphasis \\\n",
"0 2 1.0 ... 0.005563 \n",
"1 2 0.0 ... 0.006518 \n",
"2 2 1.0 ... 0.007181 \n",
"3 3 1.0 ... 0.004902 \n",
"4 2 1.0 ... 0.007222 \n",
"\n",
" original_glrlm_LongRunLowGrayLevelEmphasis \\\n",
"0 10.779989 \n",
"1 27.650685 \n",
"2 25.338218 \n",
"3 31.461354 \n",
"4 27.916261 \n",
"\n",
" original_glrlm_ShortRunHighGrayLevelEmphasis \\\n",
"0 0.789987 \n",
"1 0.442279 \n",
"2 0.503046 \n",
"3 0.399896 \n",
"4 0.473278 \n",
"\n",
" original_glszm_GrayLevelNonUniformity \\\n",
"0 27.545455 \n",
"1 78.025000 \n",
"2 72.027027 \n",
"3 99.019802 \n",
"4 56.034483 \n",
"\n",
" original_glszm_GrayLevelNonUniformityNormalized \\\n",
"0 0.834711 \n",
"1 0.975313 \n",
"2 0.973338 \n",
"3 0.980394 \n",
"4 0.966112 \n",
"\n",
" original_glszm_LargeAreaEmphasis \\\n",
"0 4067578.818 \n",
"1 2403756.075 \n",
"2 1561963.432 \n",
"3 7007670.723 \n",
"4 1288913.690 \n",
"\n",
" original_glszm_SizeZoneNonUniformityNormalized \\\n",
"0 0.180900 \n",
"1 0.198125 \n",
"2 0.275749 \n",
"3 0.253014 \n",
"4 0.216409 \n",
"\n",
" original_glszm_SmallAreaEmphasis original_ngtdm_Busyness \\\n",
"0 0.403535 473.464852 \n",
"1 0.444391 59.459710 \n",
"2 0.534549 33.935384 \n",
"3 0.506185 46.859265 \n",
"4 0.462282 39.621023 \n",
"\n",
" original_ngtdm_Strength \n",
"0 0.000758 \n",
"1 0.003685 \n",
"2 0.006447 \n",
"3 0.004543 \n",
"4 0.005626 \n",
"\n",
"[5 rows x 36 columns]"
]
},
"execution_count": 14,
"metadata": {},
"output_type": "execute_result"
}
],
"source": [
"X_corr.head()"
]
},
{
"cell_type": "code",
"execution_count": 15,
"id": "b361c4c4-54c2-4de3-b5d3-c9a276019f36",
"metadata": {},
"outputs": [],
"source": [
"# saving the feature in text file for future use in model training and prediction\n",
"with open('30cor.txt', 'w') as f:\n",
" for feature in X_corr.columns:\n",
" f.write(feature + '\\n')"
]
},
{
"cell_type": "code",
"execution_count": 16,
"id": "766f1c86-ea45-4a3d-be13-c1335f27268b",
"metadata": {},
"outputs": [],
"source": [
"from sklearn.model_selection import train_test_split\n",
"from sklearn.preprocessing import StandardScaler\n",
"# Split the correlated feature into training and testing sets ( correlated features)\n",
"X_train1, X_test1, y_train1, y_test1 = train_test_split(X_corr, y_corr, test_size=0.2,shuffle=True,random_state=42)\n",
"# Standardize the features\n",
"scaler = StandardScaler()\n",
"X_train_sc1 = scaler.fit_transform(X_train1)\n",
"X_test_sc1 = scaler.fit_transform(X_test1)"
]
},
{
"cell_type": "code",
"execution_count": 17,
"id": "790b0a2c-e36e-4f70-ac0f-a2bc8a2f30ce",
"metadata": {},
"outputs": [
{
"data": {
"text/plain": [
"((320, 36),)"
]
},
"execution_count": 17,
"metadata": {},
"output_type": "execute_result"
}
],
"source": [
"X_train1.shape, "
]
},
{
"cell_type": "code",
"execution_count": 18,
"id": "08017321-a5fe-4ae4-82c8-8f65b5b9191c",
"metadata": {},
"outputs": [],
"source": [
"# dataset using correlated feature selection for forward feature selection\n",
"X_for = df_corr.drop([\"PCR\",\"RelapseFreeSurvival (outcome)\"],axis=1)\n",
"y_for = df_corr[\"PCR\"]"
]
},
{
"cell_type": "code",
"execution_count": 19,
"id": "9e854be0-bf8e-4ad2-9d12-85584abe4f16",
"metadata": {},
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
"Selected features: Index(['ER', 'PgR', 'HER2', 'TrippleNegative', 'ChemoGrade', 'HistologyType',\n",
" 'LNStatus', 'Gene', 'original_shape_Elongation',\n",
" 'original_shape_MeshVolume', 'original_firstorder_InterquartileRange',\n",
" 'original_firstorder_Kurtosis', 'original_glcm_Imc1',\n",
" 'original_gldm_SmallDependenceEmphasis',\n",
" 'original_glrlm_LongRunLowGrayLevelEmphasis',\n",
" 'original_glszm_GrayLevelNonUniformity',\n",
" 'original_glszm_GrayLevelNonUniformityNormalized',\n",
" 'original_glszm_SizeZoneNonUniformityNormalized'],\n",
" dtype='object')\n"
]
}
],
"source": [
"import pandas as pd\n",
"from sklearn.ensemble import RandomForestClassifier\n",
"from sklearn.feature_selection import SequentialFeatureSelector\n",
"\n",
"# Handle missing values in y_for\n",
"y_for = y_for.dropna() # Alternatively, use fillna()\n",
"\n",
"# Ensure X_for matches the cleaned y_for\n",
"X_for = X_for.loc[y_for.index]\n",
"\n",
"# Sequential Feature Selector\n",
"selector = SequentialFeatureSelector(\n",
" estimator=RandomForestClassifier(n_estimators=100),\n",
" n_features_to_select='auto'\n",
")\n",
"\n",
"# Fit the selector to the data\n",
"selector.fit(X_for, y_for)\n",
"\n",
"# Get the selected features\n",
"selected_features1 = X_for.columns[selector.get_support()]\n",
"print(\"Selected features:\", selected_features1)\n"
]
},
{
"cell_type": "code",
"execution_count": 20,
"id": "662c1167-d3b1-4f56-ab3d-de46dfaf915e",
"metadata": {},
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
"<class 'pandas.core.frame.DataFrame'>\n",
"RangeIndex: 400 entries, 0 to 399\n",
"Data columns (total 18 columns):\n",
" # Column Non-Null Count Dtype \n",
"--- ------ -------------- ----- \n",
" 0 ER 400 non-null int64 \n",
" 1 PgR 400 non-null float64\n",
" 2 HER2 400 non-null float64\n",
" 3 TrippleNegative 400 non-null float64\n",
" 4 ChemoGrade 400 non-null float64\n",
" 5 HistologyType 400 non-null float64\n",
" 6 LNStatus 400 non-null float64\n",
" 7 Gene 400 non-null float64\n",
" 8 original_shape_Elongation 400 non-null float64\n",
" 9 original_shape_MeshVolume 400 non-null float64\n",
" 10 original_firstorder_InterquartileRange 400 non-null float64\n",
" 11 original_firstorder_Kurtosis 400 non-null float64\n",
" 12 original_glcm_Imc1 400 non-null float64\n",
" 13 original_gldm_SmallDependenceEmphasis 400 non-null float64\n",
" 14 original_glrlm_LongRunLowGrayLevelEmphasis 400 non-null float64\n",
" 15 original_glszm_GrayLevelNonUniformity 400 non-null float64\n",
" 16 original_glszm_GrayLevelNonUniformityNormalized 400 non-null float64\n",
" 17 original_glszm_SizeZoneNonUniformityNormalized 400 non-null float64\n",
"dtypes: float64(17), int64(1)\n",
"memory usage: 56.4 KB\n"
]
}
],
"source": [
"# Create a copy of the dataset with the selected features\n",
"X_selected_for1 = df_corr[selected_features1]\n",
"X_selected_for1.info()\n"
]
},
{
"cell_type": "code",
"execution_count": 21,
"id": "fd51f64a-7f57-493b-8485-41f3e313a61f",
"metadata": {},
"outputs": [],
"source": [
"# X and y for forward feature selection\n",
"df3 = df.copy()\n",
"X2 = df3[selected_features1]\n",
"y2 = df3[\"PCR\"]"
]
},
{
"cell_type": "code",
"execution_count": 22,
"id": "c6f0e344-3a90-4418-8f20-2646f5e66f13",
"metadata": {},
"outputs": [
{
"data": {
"text/plain": [
"(400, 18)"
]
},
"execution_count": 22,
"metadata": {},
"output_type": "execute_result"
}
],
"source": [
"X2.shape"
]
},
{
"cell_type": "code",
"execution_count": 23,
"id": "d81d0dcb-5b32-4a06-93aa-9c8d97a7d575",
"metadata": {},
"outputs": [],
"source": [
"# Splitting the data into training and testing sets ( forward feature selection)\n",
"X_train2, X_test2, y_train2, y_test2 = train_test_split(X2, y2, test_size=0.2,shuffle=True,random_state=42)\n",
"# Standardize the features\n",
"scaler = StandardScaler()\n",
"X_train_sc2 = scaler.fit_transform(X_train2)\n",
"X_test_sc2 = scaler.fit_transform(X_test2)\n"
]
},
{
"cell_type": "markdown",
"id": "ed7db500-58aa-4326-9dd4-d8c4589bac06",
"metadata": {},
"source": [
"## Models"
]
},
{
"cell_type": "markdown",
"id": "97247500-54a1-4046-bc15-4d8c2d511e97",
"metadata": {},
"source": [
"### Linear Regression "
]
},
{
"cell_type": "code",
"execution_count": 42,
"id": "b17e1da1-7a02-493f-a2a1-4907be57276b",
"metadata": {},
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
"Initial Classification Report:\n",
" precision recall f1-score support\n",
"\n",
" 0.0 0.82 0.87 0.85 94\n",
" 1.0 0.40 0.31 0.35 26\n",
"\n",
" accuracy 0.75 120\n",
" macro avg 0.61 0.59 0.60 120\n",
"weighted avg 0.73 0.75 0.74 120\n",
"\n",
"\n",
"Initial Balanced Accuracy Score:\n",
"0.5900163666121113\n"
]
}
],
"source": [
"# Import required libraries\n",
"from sklearn.linear_model import LogisticRegression\n",
"from sklearn.metrics import accuracy_score, precision_score, recall_score, f1_score, confusion_matrix, balanced_accuracy_score\n",
"\n",
"# Splitting the data into training and testing sets\n",
"X_train, X_test, y_train, y_test = train_test_split(X_corr, y_corr, test_size=0.3, shuffle=True, random_state=64)\n",
"\n",
"# Standardizing the features\n",
"scaler = StandardScaler()\n",
"X_train_norm = scaler.fit_transform(X_train)\n",
"X_test_norm = scaler.transform(X_test)\n",
"\n",
"# Initial training of Logistic Regression model (without SMOTE)\n",
"log_reg = LogisticRegression(random_state=42)\n",
"log_reg.fit(X_train_norm, y_train)\n",
"\n",
"# Predicting class labels for testing data (initial prediction)\n",
"y_pred = log_reg.predict(X_test_norm)\n",
"\n",
"# Calculating classification report and balanced accuracy score (initial performance)\n",
"report = classification_report(y_test, y_pred)\n",
"balanced_accuracy = balanced_accuracy_score(y_test, y_pred)\n",
"\n",
"print('Initial Classification Report:')\n",
"print(report)\n",
"\n",
"print('\\nInitial Balanced Accuracy Score:')\n",
"print(balanced_accuracy)\n"
]
},
{
"cell_type": "markdown",
"id": "315e1a8f-b4d6-49f1-b7a3-704178f5074f",
"metadata": {},
"source": [
"### Random forest"
]
},
{
"cell_type": "code",
"execution_count": null,
"id": "d5a51ba9-666c-4dfd-b370-a9f84555a35f",
"metadata": {},
"outputs": [],
"source": [
"from sklearn.ensemble import RandomForestClassifier\n",
"from sklearn.model_selection import GridSearchCV\n",
"from sklearn.metrics import balanced_accuracy_score, classification_report\n",
"from sklearn.impute import SimpleImputer\n",
"import pandas as pd\n",
"\n",
"# Handling missing values in features\n",
"imputer = SimpleImputer(strategy='mean')\n",
"X_train_sc1 = imputer.fit_transform(X_train_sc1)\n",
"X_train_sc2 = imputer.fit_transform(X_train_sc2)\n",
"\n",
"# Handling missing values in targets\n",
"y_train1 = pd.Series(y_train1).fillna(pd.Series(y_train1).mode()[0])\n",
"y_train2 = pd.Series(y_train2).fillna(pd.Series(y_train2).mode()[0])\n",
"\n",
"# Defining hyperparameter grid for RandomForestClassifier\n",
"param_grid = {\n",
" 'n_estimators': [100, 200, 300],\n",
" 'max_depth': [2, 5, 10],\n",
" 'min_samples_split': [2, 5, 10],\n",
" 'min_samples_leaf': [1, 2, 4],\n",
" 'random_state': [42]\n",
"}\n",
"\n",
"# Instantiate RandomForestClassifier\n",
"model1 = RandomForestClassifier()\n",
"\n",
"# Creating GridSearchCV objects for dataset1 and dataset2\n",
"grid_search1 = GridSearchCV(estimator=model1, param_grid=param_grid, scoring='balanced_accuracy', cv=5)\n",
"grid_search2 = GridSearchCV(estimator=model1, param_grid=param_grid, scoring='balanced_accuracy', cv=5)\n",
"\n",
"# Fitting the GridSearchCV objects for dataset1 and dataset2\n",
"grid_search1.fit(X_train_sc1, y_train1)\n",
"grid_search2.fit(X_train_sc2, y_train2)\n",
"\n",
"# Getting the best parameters from GridSearchCV for dataset1 and dataset2\n",
"best_params1 = grid_search1.best_params_\n",
"best_params2 = grid_search2.best_params_\n",
"\n",
"# Printing the best parameters for each dataset\n",
"print(\"Best Parameters for Dataset 1:\", best_params1)\n",
"print(\"Best Parameters for Dataset 2:\", best_params2)\n",
"\n",
"# Getting the balanced accuracy for the best model from GridSearchCV for each dataset\n",
"best_model1 = RandomForestClassifier(**best_params1)\n",
"best_model1.fit(X_train_sc1, y_train1)\n",
"y_pred1 = best_model1.predict(X_test_sc1)\n",
"balanced_accuracy1 = balanced_accuracy_score(y_test1, y_pred1)\n",
"\n",
"best_model2 = RandomForestClassifier(**best_params2)\n",
"best_model2.fit(X_train_sc2, y_train2)\n",
"y_pred2 = best_model2.predict(X_test_sc2)\n",
"balanced_accuracy2 = balanced_accuracy_score(y_test2, y_pred2)\n",
"\n",
"# Printing the balanced accuracy for each dataset\n",
"print(\"Dataset 1 Balanced Accuracy:\", balanced_accuracy1)\n",
"print(\"Dataset 2 Balanced Accuracy:\", balanced_accuracy2)\n",
"\n",
"# Getting the classification report for the best model from GridSearchCV for each dataset\n",
"print(\"Dataset 1 Classification Report:\\n\", classification_report(y_test1, y_pred1))\n",
"print(\"Dataset 2 Classification Report:\\n\", classification_report(y_test2, y_pred2))\n"
]
},
{
"cell_type": "code",
"execution_count": 34,
"id": "da95f0ab-9167-4bdf-a459-27a1a5caa8c9",
"metadata": {},
"outputs": [
{
"name": "stderr",
"output_type": "stream",
"text": [
"C:\\Users\\LLR User\\miniconda3\\envs\\MLE\\Lib\\site-packages\\sklearn\\linear_model\\_logistic.py:1197: UserWarning: l1_ratio parameter is only used when penalty is 'elasticnet'. Got (penalty=l1)\n",
" warnings.warn(\n",
"C:\\Users\\LLR User\\miniconda3\\envs\\MLE\\Lib\\site-packages\\sklearn\\linear_model\\_logistic.py:1197: UserWarning: l1_ratio parameter is only used when penalty is 'elasticnet'. Got (penalty=l1)\n",
" warnings.warn(\n",
"C:\\Users\\LLR User\\miniconda3\\envs\\MLE\\Lib\\site-packages\\sklearn\\linear_model\\_logistic.py:1197: UserWarning: l1_ratio parameter is only used when penalty is 'elasticnet'. Got (penalty=l1)\n",
" warnings.warn(\n",
"C:\\Users\\LLR User\\miniconda3\\envs\\MLE\\Lib\\site-packages\\sklearn\\linear_model\\_logistic.py:1197: UserWarning: l1_ratio parameter is only used when penalty is 'elasticnet'. Got (penalty=l1)\n",
" warnings.warn(\n",
"C:\\Users\\LLR User\\miniconda3\\envs\\MLE\\Lib\\site-packages\\sklearn\\linear_model\\_logistic.py:1197: UserWarning: l1_ratio parameter is only used when penalty is 'elasticnet'. Got (penalty=l1)\n",
" warnings.warn(\n",
"C:\\Users\\LLR User\\miniconda3\\envs\\MLE\\Lib\\site-packages\\sklearn\\linear_model\\_logistic.py:1197: UserWarning: l1_ratio parameter is only used when penalty is 'elasticnet'. Got (penalty=l2)\n",
" warnings.warn(\n",
"C:\\Users\\LLR User\\miniconda3\\envs\\MLE\\Lib\\site-packages\\sklearn\\linear_model\\_logistic.py:1197: UserWarning: l1_ratio parameter is only used when penalty is 'elasticnet'. Got (penalty=l2)\n",
" warnings.warn(\n",
"C:\\Users\\LLR User\\miniconda3\\envs\\MLE\\Lib\\site-packages\\sklearn\\linear_model\\_logistic.py:1197: UserWarning: l1_ratio parameter is only used when penalty is 'elasticnet'. Got (penalty=l2)\n",
" warnings.warn(\n",
"C:\\Users\\LLR User\\miniconda3\\envs\\MLE\\Lib\\site-packages\\sklearn\\linear_model\\_logistic.py:1197: UserWarning: l1_ratio parameter is only used when penalty is 'elasticnet'. Got (penalty=l2)\n",
" warnings.warn(\n",
"C:\\Users\\LLR User\\miniconda3\\envs\\MLE\\Lib\\site-packages\\sklearn\\linear_model\\_logistic.py:1197: UserWarning: l1_ratio parameter is only used when penalty is 'elasticnet'. Got (penalty=l2)\n",
" warnings.warn(\n",
"C:\\Users\\LLR User\\miniconda3\\envs\\MLE\\Lib\\site-packages\\sklearn\\linear_model\\_logistic.py:1197: UserWarning: l1_ratio parameter is only used when penalty is 'elasticnet'. Got (penalty=l1)\n",
" warnings.warn(\n",
"C:\\Users\\LLR User\\miniconda3\\envs\\MLE\\Lib\\site-packages\\sklearn\\linear_model\\_logistic.py:1197: UserWarning: l1_ratio parameter is only used when penalty is 'elasticnet'. Got (penalty=l1)\n",
" warnings.warn(\n",
"C:\\Users\\LLR User\\miniconda3\\envs\\MLE\\Lib\\site-packages\\sklearn\\linear_model\\_logistic.py:1197: UserWarning: l1_ratio parameter is only used when penalty is 'elasticnet'. Got (penalty=l1)\n",
" warnings.warn(\n",
"C:\\Users\\LLR User\\miniconda3\\envs\\MLE\\Lib\\site-packages\\sklearn\\linear_model\\_logistic.py:1197: UserWarning: l1_ratio parameter is only used when penalty is 'elasticnet'. Got (penalty=l1)\n",
" warnings.warn(\n",
"C:\\Users\\LLR User\\miniconda3\\envs\\MLE\\Lib\\site-packages\\sklearn\\linear_model\\_logistic.py:1197: UserWarning: l1_ratio parameter is only used when penalty is 'elasticnet'. Got (penalty=l1)\n",
" warnings.warn(\n",
"C:\\Users\\LLR User\\miniconda3\\envs\\MLE\\Lib\\site-packages\\sklearn\\linear_model\\_logistic.py:1197: UserWarning: l1_ratio parameter is only used when penalty is 'elasticnet'. Got (penalty=l2)\n",
" warnings.warn(\n",
"C:\\Users\\LLR User\\miniconda3\\envs\\MLE\\Lib\\site-packages\\sklearn\\linear_model\\_logistic.py:1197: UserWarning: l1_ratio parameter is only used when penalty is 'elasticnet'. Got (penalty=l2)\n",
" warnings.warn(\n",
"C:\\Users\\LLR User\\miniconda3\\envs\\MLE\\Lib\\site-packages\\sklearn\\linear_model\\_logistic.py:1197: UserWarning: l1_ratio parameter is only used when penalty is 'elasticnet'. Got (penalty=l2)\n",
" warnings.warn(\n",
"C:\\Users\\LLR User\\miniconda3\\envs\\MLE\\Lib\\site-packages\\sklearn\\linear_model\\_logistic.py:1197: UserWarning: l1_ratio parameter is only used when penalty is 'elasticnet'. Got (penalty=l2)\n",
" warnings.warn(\n",
"C:\\Users\\LLR User\\miniconda3\\envs\\MLE\\Lib\\site-packages\\sklearn\\linear_model\\_logistic.py:1197: UserWarning: l1_ratio parameter is only used when penalty is 'elasticnet'. Got (penalty=l2)\n",
" warnings.warn(\n",
"C:\\Users\\LLR User\\miniconda3\\envs\\MLE\\Lib\\site-packages\\sklearn\\linear_model\\_logistic.py:1197: UserWarning: l1_ratio parameter is only used when penalty is 'elasticnet'. Got (penalty=l1)\n",
" warnings.warn(\n",
"C:\\Users\\LLR User\\miniconda3\\envs\\MLE\\Lib\\site-packages\\sklearn\\linear_model\\_logistic.py:1197: UserWarning: l1_ratio parameter is only used when penalty is 'elasticnet'. Got (penalty=l1)\n",
" warnings.warn(\n",
"C:\\Users\\LLR User\\miniconda3\\envs\\MLE\\Lib\\site-packages\\sklearn\\linear_model\\_logistic.py:1197: UserWarning: l1_ratio parameter is only used when penalty is 'elasticnet'. Got (penalty=l1)\n",
" warnings.warn(\n",
"C:\\Users\\LLR User\\miniconda3\\envs\\MLE\\Lib\\site-packages\\sklearn\\linear_model\\_logistic.py:1197: UserWarning: l1_ratio parameter is only used when penalty is 'elasticnet'. Got (penalty=l1)\n",
" warnings.warn(\n",
"C:\\Users\\LLR User\\miniconda3\\envs\\MLE\\Lib\\site-packages\\sklearn\\linear_model\\_logistic.py:1197: UserWarning: l1_ratio parameter is only used when penalty is 'elasticnet'. Got (penalty=l1)\n",
" warnings.warn(\n",
"C:\\Users\\LLR User\\miniconda3\\envs\\MLE\\Lib\\site-packages\\sklearn\\linear_model\\_logistic.py:1197: UserWarning: l1_ratio parameter is only used when penalty is 'elasticnet'. Got (penalty=l2)\n",
" warnings.warn(\n",
"C:\\Users\\LLR User\\miniconda3\\envs\\MLE\\Lib\\site-packages\\sklearn\\linear_model\\_logistic.py:1197: UserWarning: l1_ratio parameter is only used when penalty is 'elasticnet'. Got (penalty=l2)\n",
" warnings.warn(\n",
"C:\\Users\\LLR User\\miniconda3\\envs\\MLE\\Lib\\site-packages\\sklearn\\linear_model\\_logistic.py:1197: UserWarning: l1_ratio parameter is only used when penalty is 'elasticnet'. Got (penalty=l2)\n",
" warnings.warn(\n",
"C:\\Users\\LLR User\\miniconda3\\envs\\MLE\\Lib\\site-packages\\sklearn\\linear_model\\_logistic.py:1197: UserWarning: l1_ratio parameter is only used when penalty is 'elasticnet'. Got (penalty=l2)\n",
" warnings.warn(\n",
"C:\\Users\\LLR User\\miniconda3\\envs\\MLE\\Lib\\site-packages\\sklearn\\linear_model\\_logistic.py:1197: UserWarning: l1_ratio parameter is only used when penalty is 'elasticnet'. Got (penalty=l2)\n",
" warnings.warn(\n",
"C:\\Users\\LLR User\\miniconda3\\envs\\MLE\\Lib\\site-packages\\sklearn\\linear_model\\_logistic.py:1197: UserWarning: l1_ratio parameter is only used when penalty is 'elasticnet'. Got (penalty=l1)\n",
" warnings.warn(\n",
"C:\\Users\\LLR User\\miniconda3\\envs\\MLE\\Lib\\site-packages\\sklearn\\linear_model\\_logistic.py:1197: UserWarning: l1_ratio parameter is only used when penalty is 'elasticnet'. Got (penalty=l1)\n",
" warnings.warn(\n",
"C:\\Users\\LLR User\\miniconda3\\envs\\MLE\\Lib\\site-packages\\sklearn\\linear_model\\_logistic.py:1197: UserWarning: l1_ratio parameter is only used when penalty is 'elasticnet'. Got (penalty=l1)\n",
" warnings.warn(\n",
"C:\\Users\\LLR User\\miniconda3\\envs\\MLE\\Lib\\site-packages\\sklearn\\linear_model\\_logistic.py:1197: UserWarning: l1_ratio parameter is only used when penalty is 'elasticnet'. Got (penalty=l1)\n",
" warnings.warn(\n",
"C:\\Users\\LLR User\\miniconda3\\envs\\MLE\\Lib\\site-packages\\sklearn\\linear_model\\_logistic.py:1197: UserWarning: l1_ratio parameter is only used when penalty is 'elasticnet'. Got (penalty=l1)\n",
" warnings.warn(\n",
"C:\\Users\\LLR User\\miniconda3\\envs\\MLE\\Lib\\site-packages\\sklearn\\linear_model\\_logistic.py:1197: UserWarning: l1_ratio parameter is only used when penalty is 'elasticnet'. Got (penalty=l2)\n",
" warnings.warn(\n",
"C:\\Users\\LLR User\\miniconda3\\envs\\MLE\\Lib\\site-packages\\sklearn\\linear_model\\_logistic.py:1197: UserWarning: l1_ratio parameter is only used when penalty is 'elasticnet'. Got (penalty=l2)\n",
" warnings.warn(\n",
"C:\\Users\\LLR User\\miniconda3\\envs\\MLE\\Lib\\site-packages\\sklearn\\linear_model\\_logistic.py:1197: UserWarning: l1_ratio parameter is only used when penalty is 'elasticnet'. Got (penalty=l2)\n",
" warnings.warn(\n",
"C:\\Users\\LLR User\\miniconda3\\envs\\MLE\\Lib\\site-packages\\sklearn\\linear_model\\_logistic.py:1197: UserWarning: l1_ratio parameter is only used when penalty is 'elasticnet'. Got (penalty=l2)\n",
" warnings.warn(\n",
"C:\\Users\\LLR User\\miniconda3\\envs\\MLE\\Lib\\site-packages\\sklearn\\linear_model\\_logistic.py:1197: UserWarning: l1_ratio parameter is only used when penalty is 'elasticnet'. Got (penalty=l2)\n",
" warnings.warn(\n",
"C:\\Users\\LLR User\\miniconda3\\envs\\MLE\\Lib\\site-packages\\sklearn\\linear_model\\_logistic.py:1197: UserWarning: l1_ratio parameter is only used when penalty is 'elasticnet'. Got (penalty=l1)\n",
" warnings.warn(\n",
"C:\\Users\\LLR User\\miniconda3\\envs\\MLE\\Lib\\site-packages\\sklearn\\linear_model\\_logistic.py:1197: UserWarning: l1_ratio parameter is only used when penalty is 'elasticnet'. Got (penalty=l1)\n",
" warnings.warn(\n",
"C:\\Users\\LLR User\\miniconda3\\envs\\MLE\\Lib\\site-packages\\sklearn\\linear_model\\_logistic.py:1197: UserWarning: l1_ratio parameter is only used when penalty is 'elasticnet'. Got (penalty=l1)\n",
" warnings.warn(\n",
"C:\\Users\\LLR User\\miniconda3\\envs\\MLE\\Lib\\site-packages\\sklearn\\linear_model\\_logistic.py:1197: UserWarning: l1_ratio parameter is only used when penalty is 'elasticnet'. Got (penalty=l1)\n",
" warnings.warn(\n",
"C:\\Users\\LLR User\\miniconda3\\envs\\MLE\\Lib\\site-packages\\sklearn\\linear_model\\_logistic.py:1197: UserWarning: l1_ratio parameter is only used when penalty is 'elasticnet'. Got (penalty=l1)\n",
" warnings.warn(\n",
"C:\\Users\\LLR User\\miniconda3\\envs\\MLE\\Lib\\site-packages\\sklearn\\linear_model\\_logistic.py:1197: UserWarning: l1_ratio parameter is only used when penalty is 'elasticnet'. Got (penalty=l2)\n",
" warnings.warn(\n",
"C:\\Users\\LLR User\\miniconda3\\envs\\MLE\\Lib\\site-packages\\sklearn\\linear_model\\_logistic.py:1197: UserWarning: l1_ratio parameter is only used when penalty is 'elasticnet'. Got (penalty=l2)\n",
" warnings.warn(\n",
"C:\\Users\\LLR User\\miniconda3\\envs\\MLE\\Lib\\site-packages\\sklearn\\linear_model\\_logistic.py:1197: UserWarning: l1_ratio parameter is only used when penalty is 'elasticnet'. Got (penalty=l2)\n",
" warnings.warn(\n",
"C:\\Users\\LLR User\\miniconda3\\envs\\MLE\\Lib\\site-packages\\sklearn\\linear_model\\_logistic.py:1197: UserWarning: l1_ratio parameter is only used when penalty is 'elasticnet'. Got (penalty=l2)\n",
" warnings.warn(\n",
"C:\\Users\\LLR User\\miniconda3\\envs\\MLE\\Lib\\site-packages\\sklearn\\linear_model\\_logistic.py:1197: UserWarning: l1_ratio parameter is only used when penalty is 'elasticnet'. Got (penalty=l2)\n",
" warnings.warn(\n",
"C:\\Users\\LLR User\\miniconda3\\envs\\MLE\\Lib\\site-packages\\sklearn\\linear_model\\_logistic.py:1197: UserWarning: l1_ratio parameter is only used when penalty is 'elasticnet'. Got (penalty=l1)\n",
" warnings.warn(\n",
"C:\\Users\\LLR User\\miniconda3\\envs\\MLE\\Lib\\site-packages\\sklearn\\linear_model\\_logistic.py:1197: UserWarning: l1_ratio parameter is only used when penalty is 'elasticnet'. Got (penalty=l1)\n",
" warnings.warn(\n",
"C:\\Users\\LLR User\\miniconda3\\envs\\MLE\\Lib\\site-packages\\sklearn\\linear_model\\_logistic.py:1197: UserWarning: l1_ratio parameter is only used when penalty is 'elasticnet'. Got (penalty=l1)\n",
" warnings.warn(\n",
"C:\\Users\\LLR User\\miniconda3\\envs\\MLE\\Lib\\site-packages\\sklearn\\linear_model\\_logistic.py:1197: UserWarning: l1_ratio parameter is only used when penalty is 'elasticnet'. Got (penalty=l1)\n",
" warnings.warn(\n",
"C:\\Users\\LLR User\\miniconda3\\envs\\MLE\\Lib\\site-packages\\sklearn\\linear_model\\_logistic.py:1197: UserWarning: l1_ratio parameter is only used when penalty is 'elasticnet'. Got (penalty=l1)\n",
" warnings.warn(\n",
"C:\\Users\\LLR User\\miniconda3\\envs\\MLE\\Lib\\site-packages\\sklearn\\linear_model\\_logistic.py:1197: UserWarning: l1_ratio parameter is only used when penalty is 'elasticnet'. Got (penalty=l2)\n",
" warnings.warn(\n",
"C:\\Users\\LLR User\\miniconda3\\envs\\MLE\\Lib\\site-packages\\sklearn\\linear_model\\_logistic.py:1197: UserWarning: l1_ratio parameter is only used when penalty is 'elasticnet'. Got (penalty=l2)\n",
" warnings.warn(\n",
"C:\\Users\\LLR User\\miniconda3\\envs\\MLE\\Lib\\site-packages\\sklearn\\linear_model\\_logistic.py:1197: UserWarning: l1_ratio parameter is only used when penalty is 'elasticnet'. Got (penalty=l2)\n",
" warnings.warn(\n",
"C:\\Users\\LLR User\\miniconda3\\envs\\MLE\\Lib\\site-packages\\sklearn\\linear_model\\_logistic.py:1197: UserWarning: l1_ratio parameter is only used when penalty is 'elasticnet'. Got (penalty=l2)\n",
" warnings.warn(\n",
"C:\\Users\\LLR User\\miniconda3\\envs\\MLE\\Lib\\site-packages\\sklearn\\linear_model\\_logistic.py:1197: UserWarning: l1_ratio parameter is only used when penalty is 'elasticnet'. Got (penalty=l2)\n",
" warnings.warn(\n",
"C:\\Users\\LLR User\\miniconda3\\envs\\MLE\\Lib\\site-packages\\sklearn\\linear_model\\_logistic.py:1197: UserWarning: l1_ratio parameter is only used when penalty is 'elasticnet'. Got (penalty=l1)\n",
" warnings.warn(\n",
"C:\\Users\\LLR User\\miniconda3\\envs\\MLE\\Lib\\site-packages\\sklearn\\linear_model\\_logistic.py:1197: UserWarning: l1_ratio parameter is only used when penalty is 'elasticnet'. Got (penalty=l1)\n",
" warnings.warn(\n",
"C:\\Users\\LLR User\\miniconda3\\envs\\MLE\\Lib\\site-packages\\sklearn\\linear_model\\_logistic.py:1197: UserWarning: l1_ratio parameter is only used when penalty is 'elasticnet'. Got (penalty=l1)\n",
" warnings.warn(\n",
"C:\\Users\\LLR User\\miniconda3\\envs\\MLE\\Lib\\site-packages\\sklearn\\linear_model\\_logistic.py:1197: UserWarning: l1_ratio parameter is only used when penalty is 'elasticnet'. Got (penalty=l1)\n",
" warnings.warn(\n",
"C:\\Users\\LLR User\\miniconda3\\envs\\MLE\\Lib\\site-packages\\sklearn\\linear_model\\_logistic.py:1197: UserWarning: l1_ratio parameter is only used when penalty is 'elasticnet'. Got (penalty=l1)\n",
" warnings.warn(\n",
"C:\\Users\\LLR User\\miniconda3\\envs\\MLE\\Lib\\site-packages\\sklearn\\linear_model\\_logistic.py:1197: UserWarning: l1_ratio parameter is only used when penalty is 'elasticnet'. Got (penalty=l2)\n",
" warnings.warn(\n",
"C:\\Users\\LLR User\\miniconda3\\envs\\MLE\\Lib\\site-packages\\sklearn\\linear_model\\_logistic.py:1197: UserWarning: l1_ratio parameter is only used when penalty is 'elasticnet'. Got (penalty=l2)\n",
" warnings.warn(\n",
"C:\\Users\\LLR User\\miniconda3\\envs\\MLE\\Lib\\site-packages\\sklearn\\linear_model\\_logistic.py:1197: UserWarning: l1_ratio parameter is only used when penalty is 'elasticnet'. Got (penalty=l2)\n",
" warnings.warn(\n",
"C:\\Users\\LLR User\\miniconda3\\envs\\MLE\\Lib\\site-packages\\sklearn\\linear_model\\_logistic.py:1197: UserWarning: l1_ratio parameter is only used when penalty is 'elasticnet'. Got (penalty=l2)\n",
" warnings.warn(\n",
"C:\\Users\\LLR User\\miniconda3\\envs\\MLE\\Lib\\site-packages\\sklearn\\linear_model\\_logistic.py:1197: UserWarning: l1_ratio parameter is only used when penalty is 'elasticnet'. Got (penalty=l2)\n",
" warnings.warn(\n",
"C:\\Users\\LLR User\\miniconda3\\envs\\MLE\\Lib\\site-packages\\sklearn\\linear_model\\_logistic.py:1197: UserWarning: l1_ratio parameter is only used when penalty is 'elasticnet'. Got (penalty=l1)\n",
" warnings.warn(\n",
"C:\\Users\\LLR User\\miniconda3\\envs\\MLE\\Lib\\site-packages\\sklearn\\linear_model\\_logistic.py:1197: UserWarning: l1_ratio parameter is only used when penalty is 'elasticnet'. Got (penalty=l1)\n",
" warnings.warn(\n",
"C:\\Users\\LLR User\\miniconda3\\envs\\MLE\\Lib\\site-packages\\sklearn\\linear_model\\_logistic.py:1197: UserWarning: l1_ratio parameter is only used when penalty is 'elasticnet'. Got (penalty=l1)\n",
" warnings.warn(\n",
"C:\\Users\\LLR User\\miniconda3\\envs\\MLE\\Lib\\site-packages\\sklearn\\linear_model\\_logistic.py:1197: UserWarning: l1_ratio parameter is only used when penalty is 'elasticnet'. Got (penalty=l1)\n",
" warnings.warn(\n",
"C:\\Users\\LLR User\\miniconda3\\envs\\MLE\\Lib\\site-packages\\sklearn\\linear_model\\_logistic.py:1197: UserWarning: l1_ratio parameter is only used when penalty is 'elasticnet'. Got (penalty=l1)\n",
" warnings.warn(\n",
"C:\\Users\\LLR User\\miniconda3\\envs\\MLE\\Lib\\site-packages\\sklearn\\linear_model\\_logistic.py:1197: UserWarning: l1_ratio parameter is only used when penalty is 'elasticnet'. Got (penalty=l2)\n",
" warnings.warn(\n",
"C:\\Users\\LLR User\\miniconda3\\envs\\MLE\\Lib\\site-packages\\sklearn\\linear_model\\_logistic.py:1197: UserWarning: l1_ratio parameter is only used when penalty is 'elasticnet'. Got (penalty=l2)\n",
" warnings.warn(\n",
"C:\\Users\\LLR User\\miniconda3\\envs\\MLE\\Lib\\site-packages\\sklearn\\linear_model\\_logistic.py:1197: UserWarning: l1_ratio parameter is only used when penalty is 'elasticnet'. Got (penalty=l2)\n",
" warnings.warn(\n",
"C:\\Users\\LLR User\\miniconda3\\envs\\MLE\\Lib\\site-packages\\sklearn\\linear_model\\_logistic.py:1197: UserWarning: l1_ratio parameter is only used when penalty is 'elasticnet'. Got (penalty=l2)\n",
" warnings.warn(\n",
"C:\\Users\\LLR User\\miniconda3\\envs\\MLE\\Lib\\site-packages\\sklearn\\linear_model\\_logistic.py:1197: UserWarning: l1_ratio parameter is only used when penalty is 'elasticnet'. Got (penalty=l2)\n",
" warnings.warn(\n",
"C:\\Users\\LLR User\\miniconda3\\envs\\MLE\\Lib\\site-packages\\sklearn\\linear_model\\_logistic.py:1197: UserWarning: l1_ratio parameter is only used when penalty is 'elasticnet'. Got (penalty=l1)\n",
" warnings.warn(\n",
"C:\\Users\\LLR User\\miniconda3\\envs\\MLE\\Lib\\site-packages\\sklearn\\linear_model\\_logistic.py:1197: UserWarning: l1_ratio parameter is only used when penalty is 'elasticnet'. Got (penalty=l1)\n",
" warnings.warn(\n",
"C:\\Users\\LLR User\\miniconda3\\envs\\MLE\\Lib\\site-packages\\sklearn\\linear_model\\_logistic.py:1197: UserWarning: l1_ratio parameter is only used when penalty is 'elasticnet'. Got (penalty=l1)\n",
" warnings.warn(\n",
"C:\\Users\\LLR User\\miniconda3\\envs\\MLE\\Lib\\site-packages\\sklearn\\linear_model\\_logistic.py:1197: UserWarning: l1_ratio parameter is only used when penalty is 'elasticnet'. Got (penalty=l1)\n",
" warnings.warn(\n",
"C:\\Users\\LLR User\\miniconda3\\envs\\MLE\\Lib\\site-packages\\sklearn\\linear_model\\_logistic.py:1197: UserWarning: l1_ratio parameter is only used when penalty is 'elasticnet'. Got (penalty=l1)\n",
" warnings.warn(\n",
"C:\\Users\\LLR User\\miniconda3\\envs\\MLE\\Lib\\site-packages\\sklearn\\linear_model\\_logistic.py:1197: UserWarning: l1_ratio parameter is only used when penalty is 'elasticnet'. Got (penalty=l2)\n",
" warnings.warn(\n",
"C:\\Users\\LLR User\\miniconda3\\envs\\MLE\\Lib\\site-packages\\sklearn\\linear_model\\_logistic.py:1197: UserWarning: l1_ratio parameter is only used when penalty is 'elasticnet'. Got (penalty=l2)\n",
" warnings.warn(\n",
"C:\\Users\\LLR User\\miniconda3\\envs\\MLE\\Lib\\site-packages\\sklearn\\linear_model\\_logistic.py:1197: UserWarning: l1_ratio parameter is only used when penalty is 'elasticnet'. Got (penalty=l2)\n",
" warnings.warn(\n",
"C:\\Users\\LLR User\\miniconda3\\envs\\MLE\\Lib\\site-packages\\sklearn\\linear_model\\_logistic.py:1197: UserWarning: l1_ratio parameter is only used when penalty is 'elasticnet'. Got (penalty=l2)\n",
" warnings.warn(\n",
"C:\\Users\\LLR User\\miniconda3\\envs\\MLE\\Lib\\site-packages\\sklearn\\linear_model\\_logistic.py:1197: UserWarning: l1_ratio parameter is only used when penalty is 'elasticnet'. Got (penalty=l2)\n",
" warnings.warn(\n",
"C:\\Users\\LLR User\\miniconda3\\envs\\MLE\\Lib\\site-packages\\sklearn\\linear_model\\_logistic.py:1197: UserWarning: l1_ratio parameter is only used when penalty is 'elasticnet'. Got (penalty=l1)\n",
" warnings.warn(\n",
"C:\\Users\\LLR User\\miniconda3\\envs\\MLE\\Lib\\site-packages\\sklearn\\linear_model\\_sag.py:349: ConvergenceWarning: The max_iter was reached which means the coef_ did not converge\n",
" warnings.warn(\n",
"C:\\Users\\LLR User\\miniconda3\\envs\\MLE\\Lib\\site-packages\\sklearn\\linear_model\\_logistic.py:1197: UserWarning: l1_ratio parameter is only used when penalty is 'elasticnet'. Got (penalty=l1)\n",
" warnings.warn(\n",
"C:\\Users\\LLR User\\miniconda3\\envs\\MLE\\Lib\\site-packages\\sklearn\\linear_model\\_sag.py:349: ConvergenceWarning: The max_iter was reached which means the coef_ did not converge\n",
" warnings.warn(\n",
"C:\\Users\\LLR User\\miniconda3\\envs\\MLE\\Lib\\site-packages\\sklearn\\linear_model\\_logistic.py:1197: UserWarning: l1_ratio parameter is only used when penalty is 'elasticnet'. Got (penalty=l1)\n",
" warnings.warn(\n",
"C:\\Users\\LLR User\\miniconda3\\envs\\MLE\\Lib\\site-packages\\sklearn\\linear_model\\_sag.py:349: ConvergenceWarning: The max_iter was reached which means the coef_ did not converge\n",
" warnings.warn(\n",
"C:\\Users\\LLR User\\miniconda3\\envs\\MLE\\Lib\\site-packages\\sklearn\\linear_model\\_logistic.py:1197: UserWarning: l1_ratio parameter is only used when penalty is 'elasticnet'. Got (penalty=l1)\n",
" warnings.warn(\n",
"C:\\Users\\LLR User\\miniconda3\\envs\\MLE\\Lib\\site-packages\\sklearn\\linear_model\\_sag.py:349: ConvergenceWarning: The max_iter was reached which means the coef_ did not converge\n",
" warnings.warn(\n",
"C:\\Users\\LLR User\\miniconda3\\envs\\MLE\\Lib\\site-packages\\sklearn\\linear_model\\_logistic.py:1197: UserWarning: l1_ratio parameter is only used when penalty is 'elasticnet'. Got (penalty=l1)\n",
" warnings.warn(\n",
"C:\\Users\\LLR User\\miniconda3\\envs\\MLE\\Lib\\site-packages\\sklearn\\linear_model\\_sag.py:349: ConvergenceWarning: The max_iter was reached which means the coef_ did not converge\n",
" warnings.warn(\n",
"C:\\Users\\LLR User\\miniconda3\\envs\\MLE\\Lib\\site-packages\\sklearn\\linear_model\\_logistic.py:1197: UserWarning: l1_ratio parameter is only used when penalty is 'elasticnet'. Got (penalty=l2)\n",
" warnings.warn(\n",
"C:\\Users\\LLR User\\miniconda3\\envs\\MLE\\Lib\\site-packages\\sklearn\\linear_model\\_sag.py:349: ConvergenceWarning: The max_iter was reached which means the coef_ did not converge\n",
" warnings.warn(\n",
"C:\\Users\\LLR User\\miniconda3\\envs\\MLE\\Lib\\site-packages\\sklearn\\linear_model\\_logistic.py:1197: UserWarning: l1_ratio parameter is only used when penalty is 'elasticnet'. Got (penalty=l2)\n",
" warnings.warn(\n",
"C:\\Users\\LLR User\\miniconda3\\envs\\MLE\\Lib\\site-packages\\sklearn\\linear_model\\_sag.py:349: ConvergenceWarning: The max_iter was reached which means the coef_ did not converge\n",
" warnings.warn(\n",
"C:\\Users\\LLR User\\miniconda3\\envs\\MLE\\Lib\\site-packages\\sklearn\\linear_model\\_logistic.py:1197: UserWarning: l1_ratio parameter is only used when penalty is 'elasticnet'. Got (penalty=l2)\n",
" warnings.warn(\n",
"C:\\Users\\LLR User\\miniconda3\\envs\\MLE\\Lib\\site-packages\\sklearn\\linear_model\\_sag.py:349: ConvergenceWarning: The max_iter was reached which means the coef_ did not converge\n",
" warnings.warn(\n",
"C:\\Users\\LLR User\\miniconda3\\envs\\MLE\\Lib\\site-packages\\sklearn\\linear_model\\_logistic.py:1197: UserWarning: l1_ratio parameter is only used when penalty is 'elasticnet'. Got (penalty=l2)\n",
" warnings.warn(\n",
"C:\\Users\\LLR User\\miniconda3\\envs\\MLE\\Lib\\site-packages\\sklearn\\linear_model\\_sag.py:349: ConvergenceWarning: The max_iter was reached which means the coef_ did not converge\n",
" warnings.warn(\n",
"C:\\Users\\LLR User\\miniconda3\\envs\\MLE\\Lib\\site-packages\\sklearn\\linear_model\\_logistic.py:1197: UserWarning: l1_ratio parameter is only used when penalty is 'elasticnet'. Got (penalty=l2)\n",
" warnings.warn(\n",
"C:\\Users\\LLR User\\miniconda3\\envs\\MLE\\Lib\\site-packages\\sklearn\\linear_model\\_sag.py:349: ConvergenceWarning: The max_iter was reached which means the coef_ did not converge\n",
" warnings.warn(\n",
"C:\\Users\\LLR User\\miniconda3\\envs\\MLE\\Lib\\site-packages\\sklearn\\linear_model\\_sag.py:349: ConvergenceWarning: The max_iter was reached which means the coef_ did not converge\n",
" warnings.warn(\n",
"C:\\Users\\LLR User\\miniconda3\\envs\\MLE\\Lib\\site-packages\\sklearn\\linear_model\\_sag.py:349: ConvergenceWarning: The max_iter was reached which means the coef_ did not converge\n",
" warnings.warn(\n",
"C:\\Users\\LLR User\\miniconda3\\envs\\MLE\\Lib\\site-packages\\sklearn\\linear_model\\_sag.py:349: ConvergenceWarning: The max_iter was reached which means the coef_ did not converge\n",
" warnings.warn(\n",
"C:\\Users\\LLR User\\miniconda3\\envs\\MLE\\Lib\\site-packages\\sklearn\\linear_model\\_sag.py:349: ConvergenceWarning: The max_iter was reached which means the coef_ did not converge\n",
" warnings.warn(\n",
"C:\\Users\\LLR User\\miniconda3\\envs\\MLE\\Lib\\site-packages\\sklearn\\linear_model\\_sag.py:349: ConvergenceWarning: The max_iter was reached which means the coef_ did not converge\n",
" warnings.warn(\n",
"C:\\Users\\LLR User\\miniconda3\\envs\\MLE\\Lib\\site-packages\\sklearn\\linear_model\\_logistic.py:1197: UserWarning: l1_ratio parameter is only used when penalty is 'elasticnet'. Got (penalty=l1)\n",
" warnings.warn(\n",
"C:\\Users\\LLR User\\miniconda3\\envs\\MLE\\Lib\\site-packages\\sklearn\\linear_model\\_sag.py:349: ConvergenceWarning: The max_iter was reached which means the coef_ did not converge\n",
" warnings.warn(\n",
"C:\\Users\\LLR User\\miniconda3\\envs\\MLE\\Lib\\site-packages\\sklearn\\linear_model\\_logistic.py:1197: UserWarning: l1_ratio parameter is only used when penalty is 'elasticnet'. Got (penalty=l1)\n",
" warnings.warn(\n",
"C:\\Users\\LLR User\\miniconda3\\envs\\MLE\\Lib\\site-packages\\sklearn\\linear_model\\_sag.py:349: ConvergenceWarning: The max_iter was reached which means the coef_ did not converge\n",
" warnings.warn(\n",
"C:\\Users\\LLR User\\miniconda3\\envs\\MLE\\Lib\\site-packages\\sklearn\\linear_model\\_logistic.py:1197: UserWarning: l1_ratio parameter is only used when penalty is 'elasticnet'. Got (penalty=l1)\n",
" warnings.warn(\n",
"C:\\Users\\LLR User\\miniconda3\\envs\\MLE\\Lib\\site-packages\\sklearn\\linear_model\\_sag.py:349: ConvergenceWarning: The max_iter was reached which means the coef_ did not converge\n",
" warnings.warn(\n",
"C:\\Users\\LLR User\\miniconda3\\envs\\MLE\\Lib\\site-packages\\sklearn\\linear_model\\_logistic.py:1197: UserWarning: l1_ratio parameter is only used when penalty is 'elasticnet'. Got (penalty=l1)\n",
" warnings.warn(\n",
"C:\\Users\\LLR User\\miniconda3\\envs\\MLE\\Lib\\site-packages\\sklearn\\linear_model\\_sag.py:349: ConvergenceWarning: The max_iter was reached which means the coef_ did not converge\n",
" warnings.warn(\n",
"C:\\Users\\LLR User\\miniconda3\\envs\\MLE\\Lib\\site-packages\\sklearn\\linear_model\\_logistic.py:1197: UserWarning: l1_ratio parameter is only used when penalty is 'elasticnet'. Got (penalty=l1)\n",
" warnings.warn(\n",
"C:\\Users\\LLR User\\miniconda3\\envs\\MLE\\Lib\\site-packages\\sklearn\\linear_model\\_sag.py:349: ConvergenceWarning: The max_iter was reached which means the coef_ did not converge\n",
" warnings.warn(\n",
"C:\\Users\\LLR User\\miniconda3\\envs\\MLE\\Lib\\site-packages\\sklearn\\linear_model\\_logistic.py:1197: UserWarning: l1_ratio parameter is only used when penalty is 'elasticnet'. Got (penalty=l2)\n",
" warnings.warn(\n",
"C:\\Users\\LLR User\\miniconda3\\envs\\MLE\\Lib\\site-packages\\sklearn\\linear_model\\_sag.py:349: ConvergenceWarning: The max_iter was reached which means the coef_ did not converge\n",
" warnings.warn(\n",
"C:\\Users\\LLR User\\miniconda3\\envs\\MLE\\Lib\\site-packages\\sklearn\\linear_model\\_logistic.py:1197: UserWarning: l1_ratio parameter is only used when penalty is 'elasticnet'. Got (penalty=l2)\n",
" warnings.warn(\n",
"C:\\Users\\LLR User\\miniconda3\\envs\\MLE\\Lib\\site-packages\\sklearn\\linear_model\\_sag.py:349: ConvergenceWarning: The max_iter was reached which means the coef_ did not converge\n",
" warnings.warn(\n",
"C:\\Users\\LLR User\\miniconda3\\envs\\MLE\\Lib\\site-packages\\sklearn\\linear_model\\_logistic.py:1197: UserWarning: l1_ratio parameter is only used when penalty is 'elasticnet'. Got (penalty=l2)\n",
" warnings.warn(\n",
"C:\\Users\\LLR User\\miniconda3\\envs\\MLE\\Lib\\site-packages\\sklearn\\linear_model\\_sag.py:349: ConvergenceWarning: The max_iter was reached which means the coef_ did not converge\n",
" warnings.warn(\n",
"C:\\Users\\LLR User\\miniconda3\\envs\\MLE\\Lib\\site-packages\\sklearn\\linear_model\\_logistic.py:1197: UserWarning: l1_ratio parameter is only used when penalty is 'elasticnet'. Got (penalty=l2)\n",
" warnings.warn(\n",
"C:\\Users\\LLR User\\miniconda3\\envs\\MLE\\Lib\\site-packages\\sklearn\\linear_model\\_sag.py:349: ConvergenceWarning: The max_iter was reached which means the coef_ did not converge\n",
" warnings.warn(\n",
"C:\\Users\\LLR User\\miniconda3\\envs\\MLE\\Lib\\site-packages\\sklearn\\linear_model\\_logistic.py:1197: UserWarning: l1_ratio parameter is only used when penalty is 'elasticnet'. Got (penalty=l2)\n",
" warnings.warn(\n",
"C:\\Users\\LLR User\\miniconda3\\envs\\MLE\\Lib\\site-packages\\sklearn\\linear_model\\_sag.py:349: ConvergenceWarning: The max_iter was reached which means the coef_ did not converge\n",
" warnings.warn(\n",
"C:\\Users\\LLR User\\miniconda3\\envs\\MLE\\Lib\\site-packages\\sklearn\\linear_model\\_sag.py:349: ConvergenceWarning: The max_iter was reached which means the coef_ did not converge\n",
" warnings.warn(\n",
"C:\\Users\\LLR User\\miniconda3\\envs\\MLE\\Lib\\site-packages\\sklearn\\linear_model\\_sag.py:349: ConvergenceWarning: The max_iter was reached which means the coef_ did not converge\n",
" warnings.warn(\n",
"C:\\Users\\LLR User\\miniconda3\\envs\\MLE\\Lib\\site-packages\\sklearn\\linear_model\\_sag.py:349: ConvergenceWarning: The max_iter was reached which means the coef_ did not converge\n",
" warnings.warn(\n",
"C:\\Users\\LLR User\\miniconda3\\envs\\MLE\\Lib\\site-packages\\sklearn\\linear_model\\_sag.py:349: ConvergenceWarning: The max_iter was reached which means the coef_ did not converge\n",
" warnings.warn(\n",
"C:\\Users\\LLR User\\miniconda3\\envs\\MLE\\Lib\\site-packages\\sklearn\\linear_model\\_sag.py:349: ConvergenceWarning: The max_iter was reached which means the coef_ did not converge\n",
" warnings.warn(\n",
"C:\\Users\\LLR User\\miniconda3\\envs\\MLE\\Lib\\site-packages\\sklearn\\linear_model\\_logistic.py:1197: UserWarning: l1_ratio parameter is only used when penalty is 'elasticnet'. Got (penalty=l1)\n",
" warnings.warn(\n",
"C:\\Users\\LLR User\\miniconda3\\envs\\MLE\\Lib\\site-packages\\sklearn\\linear_model\\_sag.py:349: ConvergenceWarning: The max_iter was reached which means the coef_ did not converge\n",
" warnings.warn(\n",
"C:\\Users\\LLR User\\miniconda3\\envs\\MLE\\Lib\\site-packages\\sklearn\\linear_model\\_logistic.py:1197: UserWarning: l1_ratio parameter is only used when penalty is 'elasticnet'. Got (penalty=l1)\n",
" warnings.warn(\n",
"C:\\Users\\LLR User\\miniconda3\\envs\\MLE\\Lib\\site-packages\\sklearn\\linear_model\\_sag.py:349: ConvergenceWarning: The max_iter was reached which means the coef_ did not converge\n",
" warnings.warn(\n",
"C:\\Users\\LLR User\\miniconda3\\envs\\MLE\\Lib\\site-packages\\sklearn\\linear_model\\_logistic.py:1197: UserWarning: l1_ratio parameter is only used when penalty is 'elasticnet'. Got (penalty=l1)\n",
" warnings.warn(\n",
"C:\\Users\\LLR User\\miniconda3\\envs\\MLE\\Lib\\site-packages\\sklearn\\linear_model\\_sag.py:349: ConvergenceWarning: The max_iter was reached which means the coef_ did not converge\n",
" warnings.warn(\n",
"C:\\Users\\LLR User\\miniconda3\\envs\\MLE\\Lib\\site-packages\\sklearn\\linear_model\\_logistic.py:1197: UserWarning: l1_ratio parameter is only used when penalty is 'elasticnet'. Got (penalty=l1)\n",
" warnings.warn(\n",
"C:\\Users\\LLR User\\miniconda3\\envs\\MLE\\Lib\\site-packages\\sklearn\\linear_model\\_sag.py:349: ConvergenceWarning: The max_iter was reached which means the coef_ did not converge\n",
" warnings.warn(\n",
"C:\\Users\\LLR User\\miniconda3\\envs\\MLE\\Lib\\site-packages\\sklearn\\linear_model\\_logistic.py:1197: UserWarning: l1_ratio parameter is only used when penalty is 'elasticnet'. Got (penalty=l1)\n",
" warnings.warn(\n",
"C:\\Users\\LLR User\\miniconda3\\envs\\MLE\\Lib\\site-packages\\sklearn\\linear_model\\_sag.py:349: ConvergenceWarning: The max_iter was reached which means the coef_ did not converge\n",
" warnings.warn(\n",
"C:\\Users\\LLR User\\miniconda3\\envs\\MLE\\Lib\\site-packages\\sklearn\\linear_model\\_logistic.py:1197: UserWarning: l1_ratio parameter is only used when penalty is 'elasticnet'. Got (penalty=l2)\n",
" warnings.warn(\n",
"C:\\Users\\LLR User\\miniconda3\\envs\\MLE\\Lib\\site-packages\\sklearn\\linear_model\\_sag.py:349: ConvergenceWarning: The max_iter was reached which means the coef_ did not converge\n",
" warnings.warn(\n",
"C:\\Users\\LLR User\\miniconda3\\envs\\MLE\\Lib\\site-packages\\sklearn\\linear_model\\_logistic.py:1197: UserWarning: l1_ratio parameter is only used when penalty is 'elasticnet'. Got (penalty=l2)\n",
" warnings.warn(\n",
"C:\\Users\\LLR User\\miniconda3\\envs\\MLE\\Lib\\site-packages\\sklearn\\linear_model\\_sag.py:349: ConvergenceWarning: The max_iter was reached which means the coef_ did not converge\n",
" warnings.warn(\n",
"C:\\Users\\LLR User\\miniconda3\\envs\\MLE\\Lib\\site-packages\\sklearn\\linear_model\\_logistic.py:1197: UserWarning: l1_ratio parameter is only used when penalty is 'elasticnet'. Got (penalty=l2)\n",
" warnings.warn(\n",
"C:\\Users\\LLR User\\miniconda3\\envs\\MLE\\Lib\\site-packages\\sklearn\\linear_model\\_sag.py:349: ConvergenceWarning: The max_iter was reached which means the coef_ did not converge\n",
" warnings.warn(\n",
"C:\\Users\\LLR User\\miniconda3\\envs\\MLE\\Lib\\site-packages\\sklearn\\linear_model\\_logistic.py:1197: UserWarning: l1_ratio parameter is only used when penalty is 'elasticnet'. Got (penalty=l2)\n",
" warnings.warn(\n",
"C:\\Users\\LLR User\\miniconda3\\envs\\MLE\\Lib\\site-packages\\sklearn\\linear_model\\_sag.py:349: ConvergenceWarning: The max_iter was reached which means the coef_ did not converge\n",
" warnings.warn(\n",
"C:\\Users\\LLR User\\miniconda3\\envs\\MLE\\Lib\\site-packages\\sklearn\\linear_model\\_logistic.py:1197: UserWarning: l1_ratio parameter is only used when penalty is 'elasticnet'. Got (penalty=l2)\n",
" warnings.warn(\n",
"C:\\Users\\LLR User\\miniconda3\\envs\\MLE\\Lib\\site-packages\\sklearn\\linear_model\\_sag.py:349: ConvergenceWarning: The max_iter was reached which means the coef_ did not converge\n",
" warnings.warn(\n",
"C:\\Users\\LLR User\\miniconda3\\envs\\MLE\\Lib\\site-packages\\sklearn\\linear_model\\_sag.py:349: ConvergenceWarning: The max_iter was reached which means the coef_ did not converge\n",
" warnings.warn(\n",
"C:\\Users\\LLR User\\miniconda3\\envs\\MLE\\Lib\\site-packages\\sklearn\\linear_model\\_sag.py:349: ConvergenceWarning: The max_iter was reached which means the coef_ did not converge\n",
" warnings.warn(\n",
"C:\\Users\\LLR User\\miniconda3\\envs\\MLE\\Lib\\site-packages\\sklearn\\linear_model\\_sag.py:349: ConvergenceWarning: The max_iter was reached which means the coef_ did not converge\n",
" warnings.warn(\n",
"C:\\Users\\LLR User\\miniconda3\\envs\\MLE\\Lib\\site-packages\\sklearn\\linear_model\\_sag.py:349: ConvergenceWarning: The max_iter was reached which means the coef_ did not converge\n",
" warnings.warn(\n",
"C:\\Users\\LLR User\\miniconda3\\envs\\MLE\\Lib\\site-packages\\sklearn\\linear_model\\_sag.py:349: ConvergenceWarning: The max_iter was reached which means the coef_ did not converge\n",
" warnings.warn(\n"
]
},
{
"name": "stdout",
"output_type": "stream",
"text": [
"Best Parameters: {'C': 1, 'l1_ratio': 0.9, 'penalty': 'elasticnet', 'solver': 'saga'}\n",
"Balanced Accuracy: 0.671875\n",
"Classification Report:\n",
" precision recall f1-score support\n",
"\n",
" 0.0 0.87 0.91 0.89 64\n",
" 1.0 0.54 0.44 0.48 16\n",
"\n",
" accuracy 0.81 80\n",
" macro avg 0.70 0.67 0.68 80\n",
"weighted avg 0.80 0.81 0.80 80\n",
"\n"
]
}
],
"source": [
"from sklearn.linear_model import LogisticRegression\n",
"from sklearn.model_selection import GridSearchCV\n",
"from sklearn.metrics import balanced_accuracy_score, classification_report\n",
"\n",
"# Simplified hyperparameter grid\n",
"param_grid = {\n",
" 'penalty': ['l1', 'l2', 'elasticnet'], # Common penalties\n",
" 'solver': ['saga'], # 'saga' supports all penalties including 'elasticnet'\n",
" 'C': [0.01, 0.1, 1, 10], # Regularization strength\n",
" 'l1_ratio': [0.1, 0.5, 0.9] # Required only for 'elasticnet'\n",
"}\n",
"\n",
"# Instantiate Logistic Regression model\n",
"model = LogisticRegression(max_iter=1000, random_state=42) # Increased max_iter for convergence\n",
"\n",
"# Create GridSearchCV\n",
"grid_search = GridSearchCV(estimator=model, param_grid=param_grid, scoring='balanced_accuracy', cv=5)\n",
"\n",
"# Fit the GridSearchCV to the data\n",
"grid_search.fit(X_train_sc1, y_train1)\n",
"\n",
"# Get the best parameters and score\n",
"best_params = grid_search.best_params_\n",
"print(\"Best Parameters:\", best_params)\n",
"\n",
"# Evaluate on the test set\n",
"best_model = LogisticRegression(**best_params, max_iter=1000, random_state=42)\n",
"best_model.fit(X_train_sc1, y_train1)\n",
"y_pred = best_model.predict(X_test_sc1)\n",
"\n",
"# Balanced accuracy and classification report\n",
"balanced_accuracy = balanced_accuracy_score(y_test1, y_pred)\n",
"print(\"Balanced Accuracy:\", balanced_accuracy)\n",
"print(\"Classification Report:\\n\", classification_report(y_test1, y_pred))\n"
]
},
{
"cell_type": "markdown",
"id": "2f818089-499c-41bc-a8e7-1f79940f18a9",
"metadata": {},
"source": [
"### SVC"
]
},
{
"cell_type": "code",
"execution_count": 27,
"id": "7e32d5dd-48fe-415a-8ab5-69a8c5716144",
"metadata": {},
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
"Dataset1: {'C': 100, 'gamma': 0.001, 'kernel': 'linear'}\n",
"Dataset2: {'C': 150, 'gamma': 0.01, 'kernel': 'rbf'}\n",
"correlation balanced accuracy: 0.640625\n",
"forward balanced accuracy: 0.546875\n",
"correlation classification report:\n",
" precision recall f1-score support\n",
"\n",
" 0.0 0.86 0.84 0.85 64\n",
" 1.0 0.41 0.44 0.42 16\n",
"\n",
" accuracy 0.76 80\n",
" macro avg 0.63 0.64 0.64 80\n",
"weighted avg 0.77 0.76 0.77 80\n",
"\n",
"forward classification report:\n",
" precision recall f1-score support\n",
"\n",
" 0.0 0.82 0.84 0.83 64\n",
" 1.0 0.29 0.25 0.27 16\n",
"\n",
" accuracy 0.72 80\n",
" macro avg 0.55 0.55 0.55 80\n",
"weighted avg 0.71 0.72 0.72 80\n",
"\n"
]
}
],
"source": [
"from sklearn.svm import SVC\n",
"from sklearn.model_selection import GridSearchCV\n",
"from sklearn.metrics import balanced_accuracy_score, classification_report\n",
"\n",
"# Defining hyperparameter grid for SVC\n",
"param_grid = {\n",
" 'kernel': ['linear', 'rbf', 'poly'],\n",
" 'C': [0.01, 0.1, 1, 10, 100, 150],\n",
" 'gamma': [0.001, 0.01, 0.1, 1]\n",
"}\n",
"\n",
"# Instantiating SVC classifier \n",
"model1 = SVC()\n",
"\n",
"# Creating the GridSearchCV objects for dataset 1 and 2 (correlation and forward selection datasets)\n",
"grid_search1 = GridSearchCV(estimator=model1, param_grid=param_grid, scoring='balanced_accuracy', cv=5)\n",
"grid_search2 = GridSearchCV(estimator=model1, param_grid=param_grid, scoring='balanced_accuracy', cv=5)\n",
"\n",
"# Fitting the GridSearchCV objects for dataset 1 and 2\n",
"grid_search1.fit(X_train_sc1, y_train1)\n",
"grid_search2.fit(X_train_sc2, y_train2)\n",
"\n",
"# Getting the best parameters from GridSearchCV for dataset 1 and 2\n",
"best_params1 = grid_search1.best_params_\n",
"best_params2 = grid_search2.best_params_\n",
"\n",
"# Printing the best parameters for each dataset\n",
"print(\"Dataset1:\", best_params1)\n",
"print(\"Dataset2:\", best_params2)\n",
"\n",
"# Getting the balanced accuracy for the best model from GridSearchCV for each dataset\n",
"best_model1 = SVC(**best_params1)\n",
"best_model1.fit(X_train_sc1, y_train1)\n",
"y_pred1 = best_model1.predict(X_test_sc1)\n",
"balanced_accuracy1 = balanced_accuracy_score(y_test1, y_pred1)\n",
"\n",
"best_model2 = SVC(**best_params2)\n",
"best_model2.fit(X_train_sc2, y_train2)\n",
"y_pred2 = best_model2.predict(X_test_sc2)\n",
"balanced_accuracy2 = balanced_accuracy_score(y_test2, y_pred2)\n",
"\n",
"# Printing the balanced accuracy for each dataset\n",
"print(\"correlation balanced accuracy:\", balanced_accuracy1)\n",
"print(\"forward balanced accuracy:\", balanced_accuracy2)\n",
"\n",
"# Getting the classification report for the best model from GridSearchCV for each dataset\n",
"print(\"correlation classification report:\\n\", classification_report(y_test1, y_pred1))\n",
"print(\"forward classification report:\\n\", classification_report(y_test2, y_pred2))\n"
]
},
{
"cell_type": "markdown",
"id": "936900ac-9890-449e-a8e7-f1d7aef72777",
"metadata": {},
"source": [
"### Decision Tree"
]
},
{
"cell_type": "code",
"execution_count": 28,
"id": "360fab3f-6fd3-488b-a6af-9432216714d1",
"metadata": {},
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
"Dataset1: {'criterion': 'gini', 'max_depth': 5, 'min_samples_leaf': 4, 'min_samples_split': 2, 'random_state': 42}\n",
"Dataset2: {'criterion': 'gini', 'max_depth': 10, 'min_samples_leaf': 1, 'min_samples_split': 5, 'random_state': 42}\n",
"correlation balanced accuracy: 0.515625\n",
"forward balanced accuracy: 0.53125\n",
"correlation classification report:\n",
" precision recall f1-score support\n",
"\n",
" 0.0 0.81 0.91 0.85 64\n",
" 1.0 0.25 0.12 0.17 16\n",
"\n",
" accuracy 0.75 80\n",
" macro avg 0.53 0.52 0.51 80\n",
"weighted avg 0.69 0.75 0.72 80\n",
"\n",
"forward classification report:\n",
" precision recall f1-score support\n",
"\n",
" 0.0 0.81 0.91 0.85 64\n",
" 1.0 0.25 0.12 0.17 16\n",
"\n",
" accuracy 0.75 80\n",
" macro avg 0.53 0.52 0.51 80\n",
"weighted avg 0.69 0.75 0.72 80\n",
"\n"
]
}
],
"source": [
"from sklearn.tree import DecisionTreeClassifier\n",
"\n",
"# Defining hyperparameter grid for DecisionTreeClassifier\n",
"param_grid = {\n",
" 'criterion': ['entropy', 'gini'],\n",
" 'max_depth': [2, 5, 10],\n",
" 'min_samples_split': [2, 5, 10],\n",
" 'min_samples_leaf': [1, 2, 4],\n",
" 'random_state': [42]\n",
"}\n",
"\n",
"# Instantiating DecisionTreeClassifier \n",
"model1 = DecisionTreeClassifier()\n",
"\n",
"# Creating GridSearchCV objects for dataset 1 and 2\n",
"grid_search1 = GridSearchCV(estimator=model1, param_grid=param_grid, scoring='balanced_accuracy', cv=5)\n",
"grid_search2 = GridSearchCV(estimator=model1, param_grid=param_grid, scoring='balanced_accuracy', cv=5)\n",
"\n",
"# Fitting the GridSearchCV objects for dataset 1 and 2\n",
"grid_search1.fit(X_train_sc1, y_train1)\n",
"grid_search2.fit(X_train_sc2, y_train2)\n",
"\n",
"# Getting the best parameters from GridSearchCV for dataset 1 and 2\n",
"best_params1 = grid_search1.best_params_\n",
"best_params2 = grid_search2.best_params_\n",
"\n",
"# Printting the best parameters for each dataset\n",
"print(\"Dataset1:\", best_params1)\n",
"print(\"Dataset2:\", best_params2)\n",
"\n",
"# Getting the balanced accuracy for the best model from GridSearchCV for each dataset\n",
"best_model1 = DecisionTreeClassifier(**best_params1)\n",
"best_model1.fit(X_train_sc1, y_train1)\n",
"y_pred1 = best_model1.predict(X_test_sc1)\n",
"balanced_accuracy1 = balanced_accuracy_score(y_test1, y_pred1)\n",
"\n",
"best_model2 = DecisionTreeClassifier(**best_params2)\n",
"best_model2.fit(X_train_sc2, y_train2)\n",
"y_pred2 = best_model2.predict(X_test_sc2)\n",
"balanced_accuracy2 = balanced_accuracy_score(y_test2, y_pred2)\n",
"\n",
"# Printting the balanced accuracy for each dataset\n",
"print(\"correlation balanced accuracy:\", balanced_accuracy1)\n",
"print(\"forward balanced accuracy:\", balanced_accuracy2)\n",
"\n",
"# Getting the classification report for the best model from GridSearchCV for each dataset\n",
"print(\"correlation classification report:\\n\", classification_report(y_test1, y_pred1))\n",
"print(\"forward classification report:\\n\", classification_report(y_test2, y_pred1))\n"
]
},
{
"cell_type": "code",
"execution_count": 29,
"id": "a26e6775-c605-4dd2-8ef2-5d89a255a8f5",
"metadata": {},
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
"Classification Report:\n",
" precision recall f1-score support\n",
"\n",
" 0.0 0.79 1.00 0.88 94\n",
" 1.0 1.00 0.04 0.07 26\n",
"\n",
" accuracy 0.79 120\n",
" macro avg 0.89 0.52 0.48 120\n",
"weighted avg 0.84 0.79 0.71 120\n",
"\n",
"\n",
"Balanced Accuracy Score:\n",
"0.5192307692307693\n"
]
}
],
"source": [
"# training the random forest model for correlated features dataset\n",
"X_train, X_test, y_train, y_test = train_test_split(X_corr, y_corr, test_size=0.3,shuffle= True, random_state=64)\n",
"scaler = StandardScaler()\n",
"X_train_norm = scaler.fit_transform(X_train)\n",
"X_test_norm = scaler.transform(X_test)\n",
"\n",
"# Creating and train Random Forest classifier with specific parameters got from gridsearch\n",
"rf_classifier = RandomForestClassifier(max_depth=10, min_samples_leaf=3, min_samples_split=15, n_estimators=1000, random_state=42)\n",
"rf_classifier.fit(X_train_norm, y_train)\n",
"\n",
"# Predicting class labels for testing data\n",
"y_pred = rf_classifier.predict(X_test_norm)\n",
"\n",
"# Calculating classification report and balanced accuracy score\n",
"classification_report = classification_report(y_test, y_pred)\n",
"balanced_accuracy = balanced_accuracy_score(y_test, y_pred)\n",
"\n",
"print('Classification Report:')\n",
"print(classification_report)\n",
"\n",
"print('\\nBalanced Accuracy Score:')\n",
"print(balanced_accuracy)\n"
]
},
{
"cell_type": "markdown",
"id": "b2491b03-0e49-4f7b-934c-300f6bfd5fa9",
"metadata": {},
"source": [
"### Smote for Random forest"
]
},
{
"cell_type": "code",
"execution_count": 30,
"id": "95e1c8ab-bec2-4eee-a4d3-ff255ff2461d",
"metadata": {},
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
"\n",
"Classification Reports for each fold:\n",
"\n",
"Fold 1 Classification Report:\n",
" precision recall f1-score support\n",
"\n",
" 0.0 0.84 0.97 0.90 33\n",
" 1.0 0.50 0.14 0.22 7\n",
"\n",
" accuracy 0.82 40\n",
" macro avg 0.67 0.56 0.56 40\n",
"weighted avg 0.78 0.82 0.78 40\n",
"\n",
"\n",
"Fold 2 Classification Report:\n",
" precision recall f1-score support\n",
"\n",
" 0.0 0.82 0.90 0.86 31\n",
" 1.0 0.50 0.33 0.40 9\n",
"\n",
" accuracy 0.78 40\n",
" macro avg 0.66 0.62 0.63 40\n",
"weighted avg 0.75 0.78 0.76 40\n",
"\n",
"\n",
"Fold 3 Classification Report:\n",
" precision recall f1-score support\n",
"\n",
" 0.0 0.79 0.97 0.87 31\n",
" 1.0 0.50 0.11 0.18 9\n",
"\n",
" accuracy 0.78 40\n",
" macro avg 0.64 0.54 0.53 40\n",
"weighted avg 0.72 0.78 0.71 40\n",
"\n",
"\n",
"Fold 4 Classification Report:\n",
" precision recall f1-score support\n",
"\n",
" 0.0 0.86 0.91 0.88 33\n",
" 1.0 0.40 0.29 0.33 7\n",
"\n",
" accuracy 0.80 40\n",
" macro avg 0.63 0.60 0.61 40\n",
"weighted avg 0.78 0.80 0.79 40\n",
"\n",
"\n",
"Fold 5 Classification Report:\n",
" precision recall f1-score support\n",
"\n",
" 0.0 0.79 0.79 0.79 29\n",
" 1.0 0.45 0.45 0.45 11\n",
"\n",
" accuracy 0.70 40\n",
" macro avg 0.62 0.62 0.62 40\n",
"weighted avg 0.70 0.70 0.70 40\n",
"\n",
"\n",
"Fold 6 Classification Report:\n",
" precision recall f1-score support\n",
"\n",
" 0.0 0.81 0.94 0.87 31\n",
" 1.0 0.50 0.22 0.31 9\n",
"\n",
" accuracy 0.78 40\n",
" macro avg 0.65 0.58 0.59 40\n",
"weighted avg 0.74 0.78 0.74 40\n",
"\n",
"\n",
"Fold 7 Classification Report:\n",
" precision recall f1-score support\n",
"\n",
" 0.0 0.78 1.00 0.88 29\n",
" 1.0 1.00 0.27 0.43 11\n",
"\n",
" accuracy 0.80 40\n",
" macro avg 0.89 0.64 0.65 40\n",
"weighted avg 0.84 0.80 0.75 40\n",
"\n",
"\n",
"Fold 8 Classification Report:\n",
" precision recall f1-score support\n",
"\n",
" 0.0 0.94 0.86 0.90 37\n",
" 1.0 0.17 0.33 0.22 3\n",
"\n",
" accuracy 0.82 40\n",
" macro avg 0.55 0.60 0.56 40\n",
"weighted avg 0.88 0.82 0.85 40\n",
"\n",
"\n",
"Fold 9 Classification Report:\n",
" precision recall f1-score support\n",
"\n",
" 0.0 0.83 0.94 0.88 32\n",
" 1.0 0.50 0.25 0.33 8\n",
"\n",
" accuracy 0.80 40\n",
" macro avg 0.67 0.59 0.61 40\n",
"weighted avg 0.77 0.80 0.77 40\n",
"\n",
"\n",
"Fold 10 Classification Report:\n",
" precision recall f1-score support\n",
"\n",
" 0.0 0.84 0.90 0.87 30\n",
" 1.0 0.62 0.50 0.56 10\n",
"\n",
" accuracy 0.80 40\n",
" macro avg 0.73 0.70 0.71 40\n",
"weighted avg 0.79 0.80 0.79 40\n",
"\n",
"\n",
"Balanced Accuracy Scores for each fold:\n",
"Fold 1 Balanced Accuracy: 0.5563\n",
"Fold 2 Balanced Accuracy: 0.6183\n",
"Fold 3 Balanced Accuracy: 0.5394\n",
"Fold 4 Balanced Accuracy: 0.5974\n",
"Fold 5 Balanced Accuracy: 0.6238\n",
"Fold 6 Balanced Accuracy: 0.5789\n",
"Fold 7 Balanced Accuracy: 0.6364\n",
"Fold 8 Balanced Accuracy: 0.5991\n",
"Fold 9 Balanced Accuracy: 0.5938\n",
"Fold 10 Balanced Accuracy: 0.7000\n",
"\n",
"Mean Balanced Accuracy Score across all folds: 0.6043275980337994\n"
]
}
],
"source": [
"from sklearn.model_selection import KFold\n",
"from imblearn.over_sampling import SMOTE\n",
"from sklearn.preprocessing import StandardScaler\n",
"from sklearn.ensemble import RandomForestClassifier\n",
"from sklearn.metrics import classification_report as clf_report, balanced_accuracy_score\n",
"import numpy as np\n",
"\n",
"# Initialize KFold with 10 splits\n",
"k = KFold(n_splits=10, shuffle=True, random_state=42)\n",
"\n",
"# Initialize SMOTE\n",
"sm = SMOTE(random_state=42)\n",
"\n",
"# Initialize Random Forest model\n",
"rf_classifier = RandomForestClassifier(max_depth=10, min_samples_leaf=3, min_samples_split=10, n_estimators=1000, random_state=42)\n",
"\n",
"# Lists to hold the results\n",
"all_classification_reports = []\n",
"all_balanced_accuracies = []\n",
"\n",
"# K-Fold Cross-Validation Loop\n",
"for train_idx, test_idx in k.split(X_corr, y_corr):\n",
" # Split the data into train and test based on the indices\n",
" X_train, X_test = X_corr.iloc[train_idx], X_corr.iloc[test_idx]\n",
" y_train, y_test = y_corr.iloc[train_idx], y_corr.iloc[test_idx]\n",
" \n",
" # Resampling the data using SMOTE only for the training data\n",
" X_train_resampled, y_train_resampled = sm.fit_resample(X_train, y_train)\n",
" \n",
" # Scaling data with StandardScaler\n",
" scaler = StandardScaler()\n",
" X_train_norm = scaler.fit_transform(X_train_resampled)\n",
" X_test_norm = scaler.transform(X_test) # Only transform test set\n",
" \n",
" # Train the Random Forest model\n",
" rf_classifier.fit(X_train_norm, y_train_resampled)\n",
" \n",
" # Predicting\n",
" y_pred = rf_classifier.predict(X_test_norm)\n",
" \n",
" # Generating the classification report and balanced accuracy score for each fold\n",
" report = clf_report(y_test, y_pred, zero_division=0)\n",
" balanced_accuracy = balanced_accuracy_score(y_test, y_pred)\n",
" \n",
" # Append results\n",
" all_classification_reports.append(report)\n",
" all_balanced_accuracies.append(balanced_accuracy)\n",
"\n",
"# Displaying the results\n",
"print(\"\\nClassification Reports for each fold:\")\n",
"for i, report in enumerate(all_classification_reports, 1):\n",
" print(f\"\\nFold {i} Classification Report:\")\n",
" print(report)\n",
"\n",
"print(\"\\nBalanced Accuracy Scores for each fold:\")\n",
"for i, score in enumerate(all_balanced_accuracies, 1):\n",
" print(f\"Fold {i} Balanced Accuracy: {score:.4f}\")\n",
"\n",
"# Optionally, you can also display the mean of the balanced accuracies\n",
"print(\"\\nMean Balanced Accuracy Score across all folds:\", np.mean(all_balanced_accuracies))\n"
]
},
{
"cell_type": "code",
"execution_count": null,
"id": "2b182154-61c7-48ec-b601-7edf2537fd0c",
"metadata": {},
"outputs": [],
"source": []
}
],
"metadata": {
"kernelspec": {
"display_name": "Python 3 (ipykernel)",
"language": "python",
"name": "python3"
},
"language_info": {
"codemirror_mode": {
"name": "ipython",
"version": 3
},
"file_extension": ".py",
"mimetype": "text/x-python",
"name": "python",
"nbconvert_exporter": "python",
"pygments_lexer": "ipython3",
"version": "3.13.0"
}
},
"nbformat": 4,
"nbformat_minor": 5
}