Subject wise LDA. subject wise scaling and PCA, renaming folders.

Results for LDA have been added to the LDA file in the comments section.
new pca outputs have been stored in the training/pca_data_v2 and testing/pca_data_v2 folders

dimensions for the pca output for each subject is not the same.

python/featureSelection.py

@@ -0,0 +1,77 @@
+import pandas as pd
+import numpy as np
+from sklearn.decomposition import PCA
+from sklearn.preprocessing import Normalizer
+
+
+
+def removeCorrelation(input_data):
+	#Columns found using analysis. Check rough-sanket.py and R code for code and results
+	cor = [24, 72]
+	vec = list(input_data.columns)
+	for c in cor:
+		vec.remove(c)
+	new_data = input_data[vec]
+	return new_data
+
+
+def runPCA(input_data, test, d):
+	input_data = removeCorrelation(input_data)
+	test = removeCorrelation(test)
+
+	normZ = Normalizer()
+	scaledX = normZ.fit_transform(input_data.iloc[:,:-1])
+	scaledTestX = normZ.transform(test)
+
+	pca = PCA()
+	pcaX = pca.fit_transform(scaledX)
+	pcaX = pd.DataFrame(pcaX)
+	print 'Approx 98% variance explained by '+str(d)+' features: ' + str(pca.explained_variance_ratio_[:d].sum())
+
+	trainY = input_data.iloc[:,-1]
+	trainY = trainY.reshape(len(trainY), 1)
+	trainY = pd.DataFrame(trainY)
+	trainY.columns = ['Class']
+
+	trainDataAfterPCA = pd.concat([pcaX.iloc[:,:d], trainY], axis=1)
+
+	testDataAfterPCA = pca.transform(scaledTestX)
+	testDataAfterPCA = pd.DataFrame(testDataAfterPCA)
+	testDataAfterPCA = testDataAfterPCA.iloc[:,:d]
+
+	return trainDataAfterPCA, testDataAfterPCA
+
+
+def mainPCA():
+	d1 = pd.read_csv('../data/Training/train_subject1_psd01.csv',header=None)
+	d2 = pd.read_csv('../data/Training/train_subject1_psd02.csv',header=None)
+	d3 = pd.read_csv('../data/Training/train_subject1_psd03.csv',header=None)
+	input_data_s1 = pd.concat([d1, d2, d3], axis=0)
+	test_s1 = pd.read_csv('../data/Testing/test_subject1_psd04.csv', header=None)
+
+	d1 = pd.read_csv('../data/Training/train_subject2_psd01.csv',header=None)
+	d2 = pd.read_csv('../data/Training/train_subject2_psd02.csv',header=None)
+	d3 = pd.read_csv('../data/Training/train_subject2_psd03.csv',header=None)
+	input_data_s2 = pd.concat([d1, d2, d3], axis=0)
+	test_s2 = pd.read_csv('../data/Testing/test_subject2_psd04.csv', header=None)
+
+	d1 = pd.read_csv('../data/Training/train_subject3_psd01.csv',header=None)
+	d2 = pd.read_csv('../data/Training/train_subject3_psd02.csv',header=None)
+	d3 = pd.read_csv('../data/Training/train_subject3_psd03.csv',header=None)
+	input_data_s3 = pd.concat([d1, d2, d3], axis=0)
+	test_s3 = pd.read_csv('../data/Testing/test_subject3_psd04.csv', header=None)
+
+	train1, test1 = runPCA(input_data_s1, test_s1, 40)
+	train2, test2 = runPCA(input_data_s2, test_s2, 50)
+	train3, test3 = runPCA(input_data_s3, test_s3, 60)
+
+	train1.to_csv('../data/Training/pca_data_v2/pca_subject1.csv', index=False)
+	train2.to_csv('../data/Training/pca_data_v2/pca_subject2.csv', index=False)
+	train3.to_csv('../data/Training/pca_data_v2/pca_subject3.csv', index=False)
+
+	test1.to_csv('../data/Testing/pca_data_v2/pca_subject1.csv', index=False)
+	test2.to_csv('../data/Testing/pca_data_v2/pca_subject2.csv', index=False)
+	test3.to_csv('../data/Testing/pca_data_v2/pca_subject3.csv', index=False)
+
+if __name__ == '__main__':
+	mainPCA()

python/linearDiscriminantAnalysis.py

@@ -0,0 +1,95 @@
+import pandas as pd
+import numpy as np
+import sys
+from sklearn.discriminant_analysis import LinearDiscriminantAnalysis, QuadraticDiscriminantAnalysis
+from sklearn.model_selection import cross_val_score
+
+
+def LDA(input_data, test, actual):
+	model = LinearDiscriminantAnalysis()
+	scores = cross_val_score(model, input_data.iloc[:,:-1], input_data.iloc[:,-1], cv = 10)
+	print 'Cross Validation Accuracy = ' + str(scores.mean())
+	model.fit(input_data.iloc[:,:-1], input_data.iloc[:,-1])
+	pred = model.predict(test)
+	accuracy = sum(pred == actual)/float(len(actual))
+	print 'Test Accuracy for the subject is = ' + str(accuracy)
+
+
+def mainRawData():
+	#Loading Input data (training)
+	d1 = pd.read_csv('../data/Training/train_subject1_psd01.csv',header=None)
+	d2 = pd.read_csv('../data/Training/train_subject1_psd02.csv',header=None)
+	d3 = pd.read_csv('../data/Training/train_subject1_psd03.csv',header=None)
+	input_data_s1 = pd.concat([d1, d2, d3], axis=0)
+
+	d1 = pd.read_csv('../data/Training/train_subject2_psd01.csv',header=None)
+	d2 = pd.read_csv('../data/Training/train_subject2_psd02.csv',header=None)
+	d3 = pd.read_csv('../data/Training/train_subject2_psd03.csv',header=None)
+	input_data_s2 = pd.concat([d1, d2, d3], axis=0)
+
+	d1 = pd.read_csv('../data/Training/train_subject3_psd01.csv',header=None)
+	d2 = pd.read_csv('../data/Training/train_subject3_psd02.csv',header=None)
+	d3 = pd.read_csv('../data/Training/train_subject3_psd03.csv',header=None)
+	input_data_s3 = pd.concat([d1, d2, d3], axis=0)
+
+	#Loading Test data (all subjects)
+	test_s1 = pd.read_csv('../data/Testing/test_subject1_psd04.csv', header=None)
+	actual_s1 = pd.read_csv('../data/Testing/ActualLables/labels_subject1_psd.csv', header=None)
+	actual_s1 = actual_s1[0]
+
+	test_s2 = pd.read_csv('../data/Testing/test_subject2_psd04.csv', header=None)
+	actual_s2 = pd.read_csv('../data/Testing/ActualLables/labels_subject2_psd.csv', header=None)
+	actual_s2 = actual_s2[0]
+
+	test_s3 = pd.read_csv('../data/Testing/test_subject3_psd04.csv', header=None)
+	actual_s3 = pd.read_csv('../data/Testing/ActualLables/labels_subject3_psd.csv', header=None)
+	actual_s3 = actual_s3[0]
+
+	LDA(input_data_s1, test_s1, actual_s1)
+	LDA(input_data_s2, test_s2, actual_s2)
+	LDA(input_data_s3, test_s3, actual_s3)
+
+	# Results:
+	#Cross Validation Accuracy = 0.69484335286
+	#Test Accuracy for the subject is = 0.714611872146
+	#
+	#Cross Validation Accuracy = 0.539059371463
+	#Test Accuracy for the subject is = 0.581221198157
+	#
+	#Cross Validation Accuracy = 0.461690546146
+	#Test Accuracy for the subject is = 0.491685779817
+
+def mainPCAData():
+	input_data_s1 = pd.read_csv('../data/Training/pca_data_v2/pca_subject1.csv')
+
+	input_data_s2 = pd.read_csv('../data/Training/pca_data_v2/pca_subject2.csv')
+
+	input_data_s3 = pd.read_csv('../data/Training/pca_data_v2/pca_subject3.csv')
+
+	#Loading Test data (all subjects)
+	test_s1 = pd.read_csv('../data/Testing/pca_data_v2/pca_subject1.csv')
+	actual_s1 = pd.read_csv('../data/Testing/ActualLables/labels_subject1_psd.csv', header=None)
+	actual_s1 = actual_s1[0]
+
+	test_s2 = pd.read_csv('../data/Testing/pca_data_v2/pca_subject2.csv')
+	actual_s2 = pd.read_csv('../data/Testing/ActualLables/labels_subject2_psd.csv', header=None)
+	actual_s2 = actual_s2[0]
+
+	test_s3 = pd.read_csv('../data/Testing/pca_data_v2/pca_subject3.csv')
+	actual_s3 = pd.read_csv('../data/Testing/ActualLables/labels_subject3_psd.csv', header=None)
+	actual_s3 = actual_s3[0]
+
+	LDA(input_data_s1, test_s1, actual_s1)
+	LDA(input_data_s2, test_s2, actual_s2)
+	LDA(input_data_s3, test_s3, actual_s3)
+
+	# Results:
+	#Cross Validation Accuracy = 0.719154247244
+	#Test Accuracy for the subject is = 0.732591324201
+	#Cross Validation Accuracy = 0.565501871535
+	#Test Accuracy for the subject is = 0.613479262673
+	#Cross Validation Accuracy = 0.482587359569
+	#Test Accuracy for the subject is = 0.505447247706
+
+if __name__ == '__main__':
+	mainPCAData()

python/rough_ankur.py

@@ -28,29 +28,7 @@
 print accuracy
 
 
-
-# Linear Discriminant Analysis
-
-#input_data = pd.read_csv('../data/')
-input_data = pd.concat([d1, d2, d3], axis=0)
-model = LinearDiscriminantAnalysis()
-scores = cross_val_score(model, input_data.iloc[:,:-1], input_data.iloc[:,-1], cv = 10)
-print scores.mean()
-
-#model.fit(input_data.iloc[:,:-1], input_data.iloc[:,-1])
-pred = model.predict(d3.iloc[:,:-1])
-actual = d3.iloc[:,-1]
-accuracy = sum(pred == actual)/float(len(actual))
-print accuracy
-
-
-d1 = pd.read_csv('../data/train_subject2_psd01.csv',header=None)
-d2 = pd.read_csv('../data/train_subject2_psd02.csv',header=None)
-d3 = pd.read_csv('../data/train_subject2_psd03.csv',header=None)
-
-d1 = pd.read_csv('../data/train_subject3_psd01.csv',header=None)
-d2 = pd.read_csv('../data/train_subject3_psd02.csv',header=None)
-d3 = pd.read_csv('../data/train_subject3_psd03.csv',header=None)
+# Random Forest
 
 
 
@@ -122,6 +100,10 @@
 from sklearn import preprocessing
 
 train = pd.read_csv('../data/Training/uncorrelated_training_data.csv')
+test1_uncor = pd.read_csv('../data/Testing/uncorrelated_subject1_data.csv')
+test2_uncor = pd.read_csv('../data/Testing/uncorrelated_subject2_data.csv')
+test3_uncor = pd.read_csv('../data/Testing/uncorrelated_subject3_data.csv')
+
 vec = list(train.columns)
 train = train[vec[:-1]]
 #preprocessing.scale(temp,axis=1,copy=False)
@@ -136,8 +118,13 @@
 test2_pca = pd.DataFrame(test2_pca)
 test3_pca = pd.DataFrame(test3_pca)
 
-test1_pca.iloc[:,:40].to_csv('../Data/Testing/pca_subject1.csv', index=False)
-test2_pca.iloc[:,:40].to_csv('../Data/Testing/pca_subject2.csv', index=False)
-test3_pca.iloc[:,:40].to_csv('../Data/Testing/pca_subject3.csv', index=False)
+test1_pca.iloc[:,:75].to_csv('../Data/Testing/pca_subject1.csv', index=False)
+test2_pca.iloc[:,:75].to_csv('../Data/Testing/pca_subject2.csv', index=False)
+test3_pca.iloc[:,:75].to_csv('../Data/Testing/pca_subject3.csv', index=False)
+
+
+
+
 
 
+#