"""

Data visualization of patient features and targets

>>> %reload_ext autoreload

>>> %autoreload 2

"""

import pandas as pd

import matplotlib.pyplot as plt

import seaborn as sns

from prescreen.vcare import targets, vital_signs

from sklearn.manifold import TSNE

from sklearn.ensemble import RandomForestClassifier

from sklearn.model_selection import train_test_split

from sklearn.metrics import roc_auc_score, roc_curve

sns.set()

df: pd.DataFrame = targets.fetch()

signs: pd.DataFrame = vital_signs.fetch()

cols_target, cols_signs = list(df.columns), list(signs.columns)

n_SF = (df['screenfail'] == 1).sum()

df_join = df.merge(signs, 'outer', on=None, left_on='id',

                        right_on='patient_id')

df_join.drop(['nip_x', 'nip_y', 'patient_id'], axis=1, inplace=True)

# pandas plot

df.plot(y='time', kind='hist', title='distribution of survival time',

        bins=128)

plt.show()

df.plot(y='status', kind='hist')

df.plot(y='screenfail', kind='hist')

plt.title('Distribution des Screen Failures')

plt.show()

print(signs.count(numeric_only=True))

# seaborn plots

sns.stripplot(x='status', y='time', data=df, jitter=True)

plt.show()

sns.countplot(x='screenfail', data=df)

plt.title('distribution of failures (n={})'.format(n_SF))

plt.show()

sns.distplot(a=signs['mean_dbp'].fillna(-1))

plt.title('mean diastelic blood pressure distribution')

plt.show()

vars= {'mean_dbp': df_join['mean_dbp'].dropna().mean(),

       'mean_sdp': df_join['mean_sdp'].dropna().mean(),

       'mean_ecog': df_join['mean_ecog'].dropna().mean(),

       'mean_height': df_join['mean_height'].dropna().mean(),

       'mean_weight': df_join['mean_weight'].dropna().mean(),

       'mean_temperature': df_join['mean_temperature'].dropna().mean(),

       'mean_eva': df_join['mean_eva'].dropna().mean(),

       'mean_sato2': df_join['mean_sato2'].dropna().mean(),

       'mean_pulse': df_join['mean_pulse'].dropna().mean(),

       'mean_frequency': df_join['mean_frequency'].dropna().mean()}

plot = sns.pairplot(data=df_join.dropna(), hue='screenfail',

             vars=list(vars.keys()), size=4)

# plot.savefig('data/pairplot_signs.png')

plt.show(plot)

X_embed = TSNE().fit_transform(df_join.fillna(value=vars)[list(vars.keys())])

df_join['embed1'] = X_embed[:, 0]

df_join['embed2'] = X_embed[:, 1]

plt.scatter(x= df_join['embed1'], y=df_join['embed2'],

            c= df_join['screenfail'],

            cmap='viridis', s=20)

plt.colorbar()

plt.show()

# df_filled = df_join.fillna(value=vars)

#

# x_train, x_test, y_train, y_test = \

#     train_test_split(df_filled[list(vars.keys())],

#                      df_join['screenfail'].fillna(0), test_size=0.33,

#                      stratify=df_join['screenfail'].fillna(0))

#

# clf = RandomForestClassifier()

# clf.fit(x_train, y_train)

#

# score = clf.score(x_test, y_test)

# auc = roc_auc_score(y_test, clf.predict_proba(x_test)[:, 1])

#

# print('accuracy: {} \n auc: {}'.format(score, auc))

#

# fpr, tpr, _ = roc_curve(y_test, clf.predict_proba(x_test)[:, 1])

#

# plt.title('baseline ROC for vital signs')

# plt.plot(fpr, tpr, 'b',

# label='AUC = %0.2f'% auc)

# plt.legend(loc='lower right')

# plt.plot([0,1],[0,1],'r--')

# plt.xlim([-0.1,1.2])

# plt.ylim([-0.1,1.2])

# plt.ylabel('True Positive Rate')

# plt.xlabel('False Positive Rate')

# plt.show()

#

#

# y_pred = clf.predict(x_test)

# y_prob = clf.predict_proba(x_test)

#

# y_pred.sum()