* implemented random forest with analysis

Author: bradenkatzman

analysis.py

@@ -35,6 +35,8 @@ def analyzeAndWriteToFile(classifier, predictions, answerKey, foldsEvaluations,
 	elif classifier.startswith("KNearestNeighbor Classifier_"):
 		idx = classifier.index("_")
 		clf = "{k}knn".format(k=classifier[idx+1:])
+	elif classifier == "Random Forest Classifier":
+		clf = "rf"
 
 
 	fileName = clf + "_" + date + "_" + t + ".txt"

knn_RNASeq.py

@@ -1,4 +1,5 @@
 from sklearn.neighbors import KNeighborsClassifier
+import sys
 import numpy as np
 
 

main.py

@@ -7,12 +7,9 @@
 import rbfSVC_RNASeq
 import neuralNetwork_RNASeq
 import knn_RNASeq
+import randomForest_RNASeq
 import analysis
 
-
-# Resource: http://machinelearningmastery.com/get-your-hands-dirty-with-scikit-learn-now/
-# Python for Java Programmers: http://python4java.necaiseweb.org/Fundamentals/TheBasics
-
 # run with down sampling and cross validation: python main.py GSE60361C13005Expression.txt expressionmRNAAnnotations.txt 1 1 1
 # run with down sampling and without cross validation: python main.py GSE60361C13005Expression.txt expressionmRNAAnnotations.txt 1 1 0
 # run without downsampling and with cross validation: python main.py GSE60361C13005Expression.txt expressionmRNAAnnotations.txt 1 0 1
@@ -54,6 +51,15 @@ def knn(trainingData, testingData, trainingDataTargets, testingDataTargets):
 
 	return knn_predictionResults
 
+def rf(trainingData, testingData, trainingDataTargets, testingDataTargets):
+	# fit training data to rf
+	randomForest_RNASeq.fitTrainingData(trainingData, trainingDataTargets)
+
+	# predict the values using random forest classifier
+	rf_predictionResults =  randomForest_RNASeq.predictTestData(testingData)
+
+	return rf_predictionResults
+
 if __name__ == '__main__':
 	t0 = time.clock()
 	print "start"
@@ -96,6 +102,11 @@ def knn(trainingData, testingData, trainingDataTargets, testingDataTargets):
 		print " - Using Multi-Layer Perceptron (Neural Network)"
 	elif classifier == 3:
 		print " - Using K Nearest Neighbor Classifier with k = {k}".format(k=n_neighbors)
+	elif classifier == 4:
+		print " - Using Random Forest Classifier"
+	else:
+		print "** ERROR: invalid classifier selection"
+		sys.exit(0)
 
 	if downSampleFlag:
 		print "** Down sampling enabled **"
@@ -197,6 +208,15 @@ def knn(trainingData, testingData, trainingDataTargets, testingDataTargets):
 					foldsEvaluations.append(analysis.calculateEvaluations(knn_predictionResults, testingDataKey))
 					# ***************** END KNN *****************
 
+				elif classifier == 4:
+					# ***************** RF *****************
+					# fit and make predictions
+					rf_predictionResults = rf(trainingFolds, testingData, trainingKeys, testingDataKey)
+
+					# add the accuracies for this fold to the accuracies list
+					foldsEvaluations.append(analysis.calculateEvaluations(rf_predictionResults, testingDataKey))
+					# ***************** END RF *****************
+
 
 				# increment iterator to process the next fold as testing data
 				iterator += 1
@@ -218,6 +238,10 @@ def knn(trainingData, testingData, trainingDataTargets, testingDataTargets):
 				# ***************** KNN *****************
 				analysis.analyzeAndWriteToFile("KNearestNeighbor Classifier_{k}".format(k=n_neighbors), knn_predictionResults, testingDataKey, foldsEvaluations, 10, 0)
 				# ***************** END KNN *****************
+			elif classifier == 4:
+				# ***************** RF *****************
+				analysis.analyzeAndWriteToFile("Random Forest Classifier", rf_predictionResults, testingDataKey, foldsEvaluations, 10, 0)
+				# ***************** END RF *****************
 
 		else:
 			# partition the down sampled data set into 70% training and 30% testing
@@ -258,7 +282,16 @@ def knn(trainingData, testingData, trainingDataTargets, testingDataTargets):
 
 				analysis.analyzeAndWriteToFile("KNearestNeighbor Classifier_{k}".format(k=n_neighbors), knn_predictionResults, data.getDSTestingDataTargetValues(), foldsEvaluations, 1, 1)
 				# ***************** END KNN *****************
+			elif classifier == 4:
+				# ***************** RF *****************
+				rf_predictionResults = rf(data.getDSTrainingData(), data.getDSTestingData(), data.getDSTargetValues(),
+					data.getDSTestingDataTargetValues())
 
+				foldsEvaluations = [] # single fold list but we still need to use a 3D list
+				foldsEvaluations.append(analysis.calculateEvaluations(rf_predictionResults, data.getDSTestingDataTargetValues()))
+
+				analysis.analyzeAndWriteToFile("Random Forest Classifier", rf_predictionResults, data.getDSTestingDataTargetValues(), foldsEvaluations, 1, 1)
+				# ***************** END RF *****************
 
 	else:
 		if crossValidateFlag:
@@ -319,6 +352,14 @@ def knn(trainingData, testingData, trainingDataTargets, testingDataTargets):
 					# add the accuracies for this fold to accuracies list
 					foldsEvaluations.append(analysis.calculateEvaluations(knn_predictionResults, testingDataKey))
 					# ***************** END KNN *****************
+				elif classifier == 4:
+					# ***************** RF *****************
+					# fit and make predictions
+					rf_predictionResults = rf(trainingFolds, testingData, trainingKeys, testingDataKey)
+
+					# add the accuracies for this fold to accuracies list
+					foldsEvaluations.append(analysis.calculateEvaluations(rf_predictionResults, testingDataKey))
+					# ***************** END RF *****************
 
 
 				# increment iterator to process the next fold as testing data
@@ -341,6 +382,10 @@ def knn(trainingData, testingData, trainingDataTargets, testingDataTargets):
 				# ***************** KNN *****************
 				analysis.analyzeAndWriteToFile("KNearestNeighbor Classifier_{k}".format(k=n_neighbors), knn_predictionResults, testingDataKey, foldsEvaluations, 10, 2)
 				# ***************** END KNN *****************
+			elif classifier == 4:
+				# ***************** RF *****************
+				analysis.analyzeAndWriteToFile("Random Forest Classifier", rf_predictionResults, testingDataKey, foldsEvaluations, 10, 2)
+				# ***************** END RF *****************
 
 		else:
 			# partition the data set into 70% training and 30% testing
@@ -386,5 +431,18 @@ def knn(trainingData, testingData, trainingDataTargets, testingDataTargets):
 				analysis.analyzeAndWriteToFile("KNearestNeighbor Classifier_{k}".format(k=n_neighbors), knn_predictionResults, data.getTestingDataTargetValues(), foldsEvaluations, 1, 3)
 				# ***************** END KNN *****************
 
+			elif classifier == 4:
+				# ***************** RF *****************
+				rf_predictionResults = rf(data.getTrainingData(), data.getTestingData(), data.getTrainingDataTargetValues(),
+					data.getTestingDataTargetValues())
+
+				# analyze results using robust evaluations
+				foldsEvaluations = []
+
+				foldsEvaluations.append(analysis.calculateEvaluations(rf_predictionResults, data.getTestingDataTargetValues()))
+
+				analysis.analyzeAndWriteToFile("Random Forest Classifier", rf_predictionResults, data.getTestingDataTargetValues(), foldsEvaluations, 1, 3)
+				# ***************** END RF *****************
+
 	print "\nprogram execution: {t} seconds".format(t=time.clock()-t0)
 	print "exiting"

randomForest_RNASeq.py

@@ -0,0 +1,26 @@
+from sklearn.ensemble import RandomForestClassifier
+import sys
+import numpy as np
+
+rfClf = RandomForestClassifier(n_estimators=10)
+
+# Source: http://scikit-learn.org/stable/modules/ensemble.html
+def fitTrainingData(training_data, nSamples):
+	# fit(X,Y)
+		# x = 2D array of cells
+		# y = 1D list of classifiers i.e. [group1, group1, group2, group2]
+
+	# makes lists into np arrays
+	training_dataNP = np.array(training_data)
+	nSamplesNP = np.array(nSamples)
+
+	rfClf.fit(training_dataNP, nSamplesNP)
+
+def predictTestData(testing_data):
+	# make list into np array
+	testing_dataNP = np.array(testing_data)
+
+	# predict the values
+	predicted = rfClf.predict(testing_dataNP)
+
+	return predicted