def run(data, labels, numFeatures, trueLabelsFile=None, verbose=False):
if (trueLabelsFile != None):
trueLabels = readLabels(trueLabelsFile)
if (verbose):
print("Selecting features...")
# feature selection
features = featureSelect(data, labels, numFeatures)
# remove columns other than selected features
newData = extractFeatures(data, features)
C_vals = [0.001, 1.0, 1000]
for C in C_vals:
# run Scikit learn SVM
predictions1 = runScikitSVM(newData, labels, C=C, verbose=verbose)
# get accuracy
if (trueLabelsFile != None):
# accuracy = calculateAccuracy(predictions1, trueLabels)
bAccuracy = calculateBalancedAccuracy(predictions1, trueLabels)
if (verbose):
# print("Accuracy:", accuracy)
print("Balanced accuracy:", bAccuracy)
# run Scikit naive bayes
predictions4 = runSciKitGaussianNaiveBayes(newData,
labels,
verbose=verbose)
# get accuracy
if (trueLabelsFile != None):
# accuracy = calculateAccuracy(predictions4, trueLabels)
bAccuracy = calculateBalancedAccuracy(predictions4, trueLabels)
if (verbose):
# print("Accuracy:", accuracy)
print("Balanced accuracy:", bAccuracy)
def test():
dataFile = "train_data/snps.data"
trueLabelsFile = "train_data/snps.labels"
labelsFilePrefix = "train_data/snps.trainlabels."
data, rows, cols = readData(dataFile, verbose=True)
listFeatures = [15, 30, 50]
# listFeatures = [100, 500]
numSplits = 10
for numFeatures in listFeatures:
for split in range(0, numSplits):
labelsFile = labelsFilePrefix + str(split)
labels = readLabels(labelsFile)
print("----------Num features:", numFeatures, "----- Labels:",
labelsFile, "----------")
run(data,
labels,
numFeatures,
trueLabelsFile=trueLabelsFile,
verbose=True)
exit()
def printPredictions(predictions):
for pred in predictions:
print(pred[0], pred[1])
if __name__ == '__main__':
if (len(sys.argv) != 4):
print("Usage:", sys.argv[0], "<dataFile> <trainLabelsFile> <k>")
exit(1)
dataFile = sys.argv[1]
labelsFile = sys.argv[2]
k_val = int(sys.argv[3])
data, _, _ = readData(dataFile)
labels = readLabels(labelsFile)
predictions = randomHyperplanes(data, labels, k_val, verbose=True, printTrainAccuracy=False)
printPredictions(predictions)
random.shuffle(allRows)
allRows = allRows[:int(numLabels * percent)]
return allRows
if __name__ == '__main__':
if (len(sys.argv) != 3):
print("Usage:", sys.argv[0], "<labelsFile> <numSplits>")
exit()
createTrainDir()
labelFile = sys.argv[1]
numSplits = int(sys.argv[2])
DATA_NAME = labelFile.split('.')[0]
OUTPUT_TRAIN_PREFIX = DATA_NAME + ".trainlabels."
splitPercent = 0.80 # 80 train 20 validation
allLabels = readLabels(labelFile)
numLabels = len(allLabels)
for i in range(numSplits):
newSplit = split(splitPercent, numLabels)
outputFileName = OUTPUT_TRAIN_PREFIX + str(i)
f = open(outputFileName, 'w')
for key in newSplit:
f.write(str(allLabels[key]) + " " + str(key) + "\n")
f.close()
if (argument == 'all'):
# datasets = ['ionosphere', 'breast_cancer', 'qsar', 'climate', 'micromass', 'hill_valley']
datasets = ['micromass', 'hill_valley']
else:
datasets = [argument]
for k in k_vals:
for datasetName in datasets:
base = '../datasets/' + datasetName + '/'
train_base = base + datasetName + '.trainlabels.'
dataFile = base + datasetName + '.data'
trueLabelsFile = base + datasetName + '.labels'
# read true labels
trueLabels = readLabels(trueLabelsFile)
C_vals = [0.001, 0.01, 0.1, 1, 10, 100]
# read data
data, _, _ = readData(dataFile)
for split in range(numSplits):
print("----------------------------------------------")
print(datasetName, "split", split)
labelsFile = train_base + str(split)
# read labels
labels = readLabels(labelsFile)
initial = True
# run regular SVM