Added ML Models for future Integration

Author: ashpande

images/download.jpg

@@ -0,0 +1,394 @@
+#!/usr/bin/env python
+# coding: utf-8
+
+# # import all depedencies
+
+# In[1]:
+
+
+import numpy as np
+import pandas as pd
+import matplotlib.pyplot as plt
+from time import sleep
+import os
+import pyeeg
+import warnings
+warnings.filterwarnings("ignore", category=np.VisibleDeprecationWarning)
+# get_ipython().run_line_magic('matplotlib', 'notebook')
+import joblib
+import pickle
+
+# # Load files from folders
+
+# In[2]:
+
+
+# import each and every file
+
+dirB="./Datasets/setB/"
+tempB = []
+for file in os.listdir(dirB):
+    fl = dirB + file
+    tempB.append(fl)
+tempB = sorted(tempB)    # class: 1     val:  -1
+
+dirC="./Datasets/setC/"
+tempC = []
+for file in os.listdir(dirC):
+    fl = dirC + file
+    tempC.append(fl)
+tempC = sorted(tempC)   # class:2       val: 0
+
+dirE="./Datasets/setE/"
+tempE = []
+for file in os.listdir(dirE):
+    fl = dirE + file
+    tempE.append(fl)
+tempE = sorted(tempE)   # class: 3        val: 1
+
+
+# # create small tables
+
+# In[3]:
+
+
+tb=[]
+st = 'A'
+for i in range(len(tempB)):
+    x = pd.read_table(tempB[i],header=None)
+    x.columns=[st+str(i)]
+    tb.append(x)
+
+tc=[]
+st = 'A'
+for i in range(len(tempC)):
+    x = pd.read_table(tempC[i],header=None)
+    x.columns=[st+str(i)]
+    tc.append(x)
+
+te=[]
+st = 'A'
+for i in range(len(tempE)):
+    x = pd.read_table(tempE[i],header=None)
+    x.columns=[st+str(i)]
+    te.append(x)
+
+
+# # make big table
+
+# In[4]:
+
+
+def table(table):
+    big_table = None
+    for ta in table:
+        big_table = pd.concat([big_table, ta],axis=1)
+    return big_table
+
+
+# In[5]:
+
+
+bigB = table(tb)
+bigC = table(tc)
+bigE = table(te)
+head = list(bigB.columns.values)
+
+
+# In[6]:
+
+
+len(bigB.columns)
+
+
+# In[7]:
+
+
+bigB.head(10)
+
+
+# # creat sub matrix for each
+
+# In[8]:
+
+
+# creat a matrix
+def creat_mat(mat):
+    matx = np.zeros((len(mat),(len(head))))
+    for i in range(len(head)):
+        matx[:,i] = mat[head[i]]
+        sleep(0.01)
+    return matx
+# every mat contain probs mat[:,i] total 28 start form mat1
+
+
+# In[9]:
+
+
+matB = creat_mat(bigB) # : refers to healthy data
+matC = creat_mat(bigC) # : refers to Inter-ictal (transition between healthy to seizure)
+matE = creat_mat(bigE) # : of ictal or seizures
+
+matB = np.nan_to_num(matB) # matB[:,0] --- > channel 0, matB[:,1] --- > channel 1 like that
+matC = np.nan_to_num(matC)
+matE = np.nan_to_num(matE)
+
+
+# 4097 data point per channel
+# 173.61 Hz sample rate and there are 4097 data point for each channel
+# total 100 channel are their
+# 4097/173.61 = 23.59 sec
+# the raw data from one of the channels for the 23.59 sec
+
+
+# # Visualize the Datapoints
+
+# In[10]:
+
+
+hl,      = plt.plot(matB[0],label='healthy')
+trans,   = plt.plot(matC[0],label='Inter-ictal')
+seizure, = plt.plot(matE[0],label='seizures')
+plt.legend(handles=[hl,trans,seizure])
+plt.savefig("fig1.png")
+
+
+# #  <---------  create number of featuters ------------>
+
+# # (DFA,HFD,SVD_Entropy,Fisher_Information,PFD)
+
+# In[11]:
+
+
+help(pyeeg.dfa),help(pyeeg.hfd),help(pyeeg.svd_entropy),help(pyeeg.fisher_info),help(pyeeg.pfd)
+
+
+# In[12]:
+
+
+# source: https://www.hindawi.com/journals/cin/2011/406391/
+def features(mat):
+    Kmax = 5
+    Tau  = 4
+    DE   = 10
+    M    = 10
+    R    = 0.3
+    Band = np.arange(1,86)
+    Fs   = 173
+    DFA                = pyeeg.dfa(mat)
+    HFD                = pyeeg.hfd(mat,Kmax)
+    SVD_Entropy        = pyeeg.svd_entropy(mat,Tau,DE)
+    Fisher_Information = pyeeg.fisher_info(mat,Tau,DE)
+    PFD                = pyeeg.pfd(mat)
+    sleep(0.01)
+
+    return (DFA,HFD,SVD_Entropy,Fisher_Information,PFD)
+
+
+# # Compute the features of datasets
+
+# In[13]:
+
+
+'''
+# this is three classification proble
+# class1: healthy       +1
+# class2: transition     0
+# class3: dieases       -1
+
+'''
+
+
+# In[14]:
+
+
+# create features of class I
+f1_B    = np.zeros((100,1))
+f2_B    = np.zeros((100,1))
+f3_B    = np.zeros((100,1))
+f4_B    = np.zeros((100,1))
+f5_B    = np.zeros((100,1))
+cl_B    = np.ones((100,1)) # 1
+
+
+for i in range(100):
+    [f1_B[i,0],f2_B[i,0],f3_B[i,0],f4_B[i,0],f5_B[i,0]]=features(matB[:,i])
+
+
+
+# In[15]:
+
+
+# create features of class II
+f1_C    = np.zeros((100,1))
+f2_C    = np.zeros((100,1))
+f3_C    = np.zeros((100,1))
+f4_C    = np.zeros((100,1))
+f5_C    = np.zeros((100,1))
+cl_C    = np.zeros((100,1)) # transition means 0
+
+for i in range(100):
+    [f1_C[i,0],f2_C[i,0],f3_C[i,0],f4_C[i,0],f5_C[i,0]]=features(matC[:,i])
+
+
+
+# In[16]:
+
+
+# create features of class III
+f1_E    = np.zeros((100,1))
+f2_E    = np.zeros((100,1))
+f3_E    = np.zeros((100,1))
+f4_E    = np.zeros((100,1))
+f5_E    = np.zeros((100,1))
+cl_E    = np.negative(np.ones((100,1))) # -1
+
+for i in range(100):
+    [f1_E[i,0],f2_E[i,0],f3_E[i,0],f4_E[i,0],f5_E[i,0]]=features(matE[:,i])
+
+
+
+# # create Feature and class matrix for each Classes
+
+# In[17]:
+
+
+MftB = np.concatenate([f1_B,f2_B,f3_B,f4_B,f5_B,cl_B],axis=1)
+MftC = np.concatenate([f1_C,f2_C,f3_C,f4_C,f5_C,cl_C],axis=1)
+MftE = np.concatenate([f1_E,f2_E,f3_E,f4_E,f5_E,cl_E],axis=1)
+
+
+# In[18]:
+
+
+FCM_B = pd.DataFrame(MftB,columns=['f1','f2','f3','f4','f5','class'])
+FCM_C = pd.DataFrame(MftC,columns=['f1','f2','f3','f4','f5','class'])
+FCM_E = pd.DataFrame(MftE,columns=['f1','f2','f3','f4','f5','class'])
+FCM_B.head(4)
+
+
+# In[19]:
+
+
+TotalDataset = pd.concat([FCM_B,FCM_C,FCM_E],ignore_index=True)
+visDat = TotalDataset.copy(deep=True)
+visDat['class'] = visDat['class'].map({1:'healthy',0:'transation',-1:'seizure'})
+
+
+# In[20]:
+
+
+import seaborn as sbn
+sbn.set(style="whitegrid", palette="muted")
+
+
+# In[21]:
+
+
+visDat.head(5)
+
+
+# In[22]:
+
+
+sbn.pairplot(visDat,hue='class',palette="husl")
+plt.savefig("fig2.png")
+
+
+# In[23]:
+
+
+plt.plot(visDat['f1'],'--o')
+
+
+# # Split arrays or matrices into random train and test subsets
+
+# In[24]:
+
+
+from sklearn.model_selection import train_test_split
+
+
+# In[25]:
+
+
+X = TotalDataset[['f1','f2','f3','f4','f5']]
+y = TotalDataset[['class']]
+X = np.asarray(X)
+y = np.asarray(y)
+
+
+# In[26]:
+
+
+X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.33, random_state=42)
+
+
+# # apply machine learning
+
+# In[27]:
+
+
+from sklearn.model_selection import train_test_split
+from sklearn.preprocessing import StandardScaler
+from sklearn.datasets import make_moons, make_circles, make_classification
+from sklearn.neural_network import MLPClassifier
+from sklearn.neighbors import KNeighborsClassifier
+from sklearn.svm import SVC
+from sklearn.gaussian_process import GaussianProcessClassifier
+from sklearn.gaussian_process.kernels import RBF
+from sklearn.tree import DecisionTreeClassifier
+from sklearn.ensemble import RandomForestClassifier, AdaBoostClassifier
+from sklearn.naive_bayes import GaussianNB
+from sklearn.discriminant_analysis import QuadraticDiscriminantAnalysis
+
+
+# In[28]:
+
+
+names = ["Nearest Neighbors", "Linear SVM", "Gaussian Process",
+         "Decision Tree", "Random Forest", "AdaBoost",
+         "Naive Bayes"]
+
+
+# In[29]:
+
+
+classifiers = [
+    KNeighborsClassifier(3),
+    SVC(kernel="linear", C=0.025),
+    GaussianProcessClassifier(1.0 * RBF(1.0), warm_start=True),
+    DecisionTreeClassifier(max_depth=5),
+    RandomForestClassifier(max_depth=5, n_estimators=10, max_features=1),
+    AdaBoostClassifier(),
+    GaussianNB()]
+
+
+# In[37]:
+
+
+clf_score=[]
+with warnings.catch_warnings():
+    warnings.simplefilter("ignore")
+    for name, clf in zip(names, classifiers):
+        clf.fit(X_train, y_train)
+        if(name=="Gaussian Process"):
+            gaussian = pickle.dumps(clf)
+        score = clf.score(X_test, y_test)
+        clf_score.append([score,name])
+
+
+# In[31]:
+
+
+clf_score
+
+
+# In[42]:
+
+
+# print(gaussian)
+joblib.dump(gaussian,'eegmodel.sav')
+
+# eegmodel =
+pickle.dump(gaussian, open('eegmodel.sav', 'wb'))