import os
import sys
import pywt
from pywt import wavedec
from __init__ import ap_entropy, samp_entropy
import numpy as np
import matplotlib
from matplotlib import pyplot as plt
from mnist_new1 import mlp, create_training_set
from pylab import *
A=[]
B=[]
C=[]
D=[]
E=[]
for fl in os.listdir("../../../ALL/A/"):
inp = []
path = "../../../ALL/A/" + fl
txt = open(path,'r')
for line in txt:
feature = line.split()[0]
inp.append(feature)
a = np.array(inp)
A.append(a)
for fl in os.listdir("../../../ALL/B/"):
inp = []
path = "../../../ALL/B/" + fl
txt = open(path,'r')
for line in txt:
feature = line.split()[0]
inp.append(feature)
a = np.array(inp)
B.append(a)
for fl in os.listdir("../../../ALL/C/"):
inp = []
path = "../../../ALL/C/" + fl
txt = open(path,'r')
for line in txt:
feature = line.split()[0]
inp.append(feature)
a = np.array(inp)
C.append(a)
for fl in os.listdir("../../../ALL/D/"):
inp = []
path = "../../../ALL/D/" + fl
txt = open(path,'r')
for line in txt:
feature = line.split()[0]
inp.append(feature)
a = np.array(inp)
D.append(a)
for fl in os.listdir("../../../ALL/E/"):
inp = []
path = "../../../ALL/E/" + fl
txt = open(path,'r')
for line in txt:
feature = line.split()[0]
inp.append(feature)
a = np.array(inp)
E.append(a)
A_ = []
B_ = []
C_ = []
D_ = []
E_ = []
for x in A:
coeffs = wavedec(x,'db4',level=8)
A_.append(coeffs)
for x in B:
coeffs = wavedec(x,'db4',level=8)
B_.append(coeffs)
for x in C:
coeffs = wavedec(x,'db4',level=8)
C_.append(coeffs)
for x in D:
coeffs = wavedec(x,'db4',level=8)
D_.append(coeffs)
for x in E:
coeffs = wavedec(x,'db4',level=8)
E_.append(coeffs)
a=[]
b=[]
c=[]
d=[]
e=[]
y_a = []
y_b = []
y_c = []
y_d = []
y_e = []
f=[]
y_f=[]
minm = [1000000000000 for i in range(1,28)]
maxm = [0 for i in range(1,28)]
inp = []
out = []
for x in A_:
features = []
j=0
for y in x:
coef = np.array(y)
energy = np.sum(coef**2)
minm[j] = min(minm[j],energy)
maxm[j] = max(maxm[j],energy)
j=j+1
#approx_en = ap_entropy(coef,2,0.5)
#minm[j] = min(minm[j],approx_en)
#maxm[j] = max(maxm[j],approx_en)
#j=j+1
#samp_en = samp_entropy(coef,2,0.5)
#minm[j] = min(minm[j],samp_en)
#maxm[j] = max(maxm[j],samp_en)
#j=j+1
mean = np.mean(coef)
minm[j]= min(minm[j],mean)
maxm[j] = max(maxm[j],mean)
j=j+1
std = np.std(coef)
minm[j] = min(minm[j],std)
maxm[j] = max(maxm[j],std)
j=j+1
features.append(energy)
#features.append(approx_en)
#features.append(samp_en)
features.append(mean)
features.append(std)
a.append(features)
y_a.append(0)
print("A done")
for x in B_:
features = []
j=0
for y in x:
coef = np.array(y)
energy = np.sum(coef**2)
minm[j] = min(minm[j],energy)
maxm[j] = max(maxm[j],energy)
j=j+1
#approx_en = ap_entropy(coef,2,0.5)
#minm[j] = min(minm[j],approx_en)
#maxm[j] = max(maxm[j],approx_en)
#j=j+1
#samp_en = samp_entropy(coef,2,0.5)
#minm[j] = min(minm[j],samp_en)
#maxm[j] = max(maxm[j],samp_en)
#j=j+1
mean = np.mean(coef)
minm[j]= min(minm[j],mean)
maxm[j] = max(maxm[j],mean)
j=j+1
std = np.std(coef)
minm[j] = min(minm[j],std)
maxm[j]= max(maxm[j],std)
j=j+1
features.append(energy)
#features.append(approx_en)
#features.append(samp_en)
features.append(mean)
features.append(std)
b.append(features)
y_b.append(0)
print("B done")
for x in C_:
features = []
j=0
for y in x:
coef = np.array(y)
energy = np.sum(coef**2)
minm[j] = min(minm[j],energy)
maxm[j] = max(maxm[j],energy)
j=j+1
#approx_en = ap_entropy(coef,2,0.5)
#minm[j] = min(minm[j],approx_en)
#maxm[j] = max(maxm[j],approx_en)
#j=j+1
#samp_en = samp_entropy(coef,2,0.5)
#minm[j] = min(minm[j],samp_en)
#maxm[j] = max(maxm[j],samp_en)
#j=j+1
mean = np.mean(coef)
minm[j]= min(minm[j],mean)
maxm[j] = max(maxm[j],mean)
j=j+1
std = np.std(coef)
minm[j] = min(minm[j],std)
maxm[j]= max(maxm[j],std)
j=j+1
features.append(energy)
#features.append(approx_en)
#features.append(samp_en)
features.append(mean)
features.append(std)
c.append(features)
y_c.append(1)
print("C done")
for x in D_:
features = []
j=0
for y in x:
coef = np.array(y)
energy = np.sum(coef**2)
minm[j] = min(minm[j],energy)
maxm[j] = max(maxm[j],energy)
j=j+1
#approx_en = ap_entropy(coef,2,0.5)
#minm[j] = min(minm[j],approx_en)
#maxm[j] = max(maxm[j],approx_en)
#j=j+1
#samp_en = samp_entropy(coef,2,0.5)
#minm[j] = min(minm[j],samp_en)
#maxm[j] = max(maxm[j],samp_en)
#j=j+1
mean = np.mean(coef)
minm[j]= min(minm[j],mean)
maxm[j] = max(maxm[j],mean)
j=j+1
std = np.std(coef)
minm[j] = min(minm[j],std)
maxm[j]= max(maxm[j],std)
j=j+1
features.append(energy)
#features.append(approx_en)
#features.append(samp_en)
features.append(mean)
features.append(std)
d.append(features)
y_d.append(1)
print("D Done")
for x in E_:
features = []
j=0
for y in x:
coef = np.array(y)
energy = np.sum(coef**2)
minm[j] = min(minm[j],energy)
maxm[j] = max(maxm[j],energy)
j=j+1
#approx_en = ap_entropy(coef,2,0.5)
#minm[j] = min(minm[j],approx_en)
#maxm[j] = max(maxm[j],approx_en)
#j=j+1
#samp_en = samp_entropy(coef,2,0.5)
#minm[j] = min(minm[j],samp_en)
#maxm[j] = max(maxm[j],samp_en)
#j=j+1
mean = np.mean(coef)
minm[j]= min(minm[j],mean)
maxm[j] = max(maxm[j],mean)
j=j+1
std = np.std(coef)
minm[j] = min(minm[j],std)
maxm[j]= max(maxm[j],std)
j=j+1
features.append(energy)
#features.append(approx_en)
#features.append(samp_en)
features.append(mean)
features.append(std)
e.append(features)
f.append(features)
y_e.append(2)
y_f.append(2)
print("E done")
i=0
while i < 100:
inp.append(a[i])
out.append(y_a[i])
inp.append(b[i])
out.append(y_b[i])
inp.append(c[i])
out.append(y_c[i])
inp.append(d[i])
out.append(y_d[i])
inp.append(e[i])
out.append(y_e[i])
inp.append(f[i])
out.append(y_f[i])
i=i+1
print("done")
i=0
z= [i for i in range(1,len(e)+1)]
import matplotlib.pyplot as plt
i=0
while i < 27:
j=0
p=[]
q=[]
r=[]
s=[]
t=[]
u=[]
while j < len(e):
p.append(a[j][i])
#q.append(b[j][i])
r.append(c[j][i])
#s.append(d[j][i])
t.append(e[j][i])
j=j+1
plt.plot(z,np.array(p), label="normal")
#plt.plot(z,np.array(q))
plt.plot(z,np.array(r), label="ephilepsy without seizures")
#plt.plot(z,np.array(s))
plt.plot(z,np.array(t), label="ephilepsy with seizures")
legend()
plt.savefig("images/feature_"+str(i)+"_out.jpg")
print("saved")
plt.clf()
i=i+1
from sklearn.ensemble import ExtraTreesClassifier
i=0
for x in inp:
j=0
while j < 27:
den = maxm[j]-minm[j]
if(maxm[j]-minm[j]==0):
den=np.amax(np.array(maxm))
inp[i][j] = np.float32((inp[i][j]-minm[j]))/np.float32(den)
if(inp[i][j]<0.00001):
inp[i][j]=0.0001
if(inp[i][j]>0.99999):
inp[i][j]=0.9999
inp[i][j] = np.float32(inp[i][j])
j=j+1
i=i+1
inp = np.array(inp)
out = np.array(out)
forest = ExtraTreesClassifier(n_estimators=100,random_state=0)
forest.fit(inp,out)
importances = forest.feature_importances_
std = np.std([tree.feature_importances_ for tree in forest.estimators_],
axis=0)
indices = np.argsort(importances)[::-1]
print("Feature ranking:")
for f in range(inp.shape[1]):
print("%d. feature %d (%f)" % (f + 1, indices[f], importances[indices[f]]))
plt.figure()
plt.title("Feature importances")
plt.bar(range(inp.shape[1]), importances[indices],
color="r", yerr=std[indices], align="center")
plt.xticks(range(inp.shape[1]), indices)
plt.xlim([-1, inp.shape[1]])
plt.savefig("images/feature_importance_out.jpg")
plt.clf()
Datasets
Models
