bestROC = -numpy.Inf
bestROC50 = -numpy.Inf
bestSuccessRate = None
bestW = None
bestK1 = 0
bestK2 = 0
K1 = [8, 9, 10, 11, 12, 13]
K2 = [8, 9, 10, 11, 12, 13]
for k1 in K1:
for k2 in [x for x in K2 if x >= k1]:
for C in Cs:
print "**** Train/Test with K1: " + str(k1) + ", k2: " + str(k2) + ", C: " + str(C)
trainData = generate_model.get_spectrum_data(
trainSeqFile, k1, k2, trainPosLen, trainNegLen, normalize=True, repeatCount=True
)
s = svm.SVM(C=C)
s.train(trainData)
results = s.stratifiedCV(trainData, numFolds=numFolds, seen=1)
labels = results.getGivenClass()
roc = results.getROC()
roc50 = results.getROCn()
successRate = results.getSuccessRate()
print "ROC: " + str(roc) + ", ROC50: " + str(roc50) + ", Success Rate: " + str(successRate)
if roc > bestROC:
# Delete the old data file.
os.remove(dataFile) if os.path.exists(dataFile) else None
# Save the training data for K1, K2 combination.
trainData.save(dataFile)
confMap = ReadConfFile(sys.argv[1])
dataFile = confMap["data"]["dataFile"]
k1 = int(confMap["kspectrum"]["k1"])
k2 = int(confMap["kspectrum"]["k2"])
C = float(confMap["kspectrum"]["C"])
posLen = int(confMap["data"]["posLen"])
negLen = int(confMap["data"]["negLen"])
numFolds = int(confMap["kspectrum"]["folds"])
featureFile = confMap["output"]["featureFile"]
resultFile = confMap["output"]["resultsFile"]
modelFile = confMap["output"]["modelFile"]
buildModel = False
trainData = generate_model.get_spectrum_data(dataFile,
k1,
k2,
posLen,
negLen,
normalize=True,
repeatCount=True)
if buildModel:
results, m = BuildModel(trainData, modelFile, numFolds)
print_results(results, resultFile, k1, k2)
find_features(trainData, featureFile, m.classifier.C)
else:
results = CrossValidate(trainData, numFolds)
print_results(results, resultFile, k1, k2)
find_features(trainData, featureFile)
outFeatureFile = sys.argv[5]
bestC = None
bestAUC = -numpy.Inf
bestFP = None
bestTP = None
K1 = [7, 8, 9, 10, 11, 12, 13]
K2 = [7, 8, 9, 10, 11, 12, 13]
result_file = open("K-spectrum.txt", "w")
for k1 in K1:
for k2 in K2:
for C in Cs:
print "**** Train/Test with K1: " + str(k1) + ", k2: " + str(k2) + ", C: " + str(C)
trainData = generate_model.get_spectrum_data(trainSeqFile, k1, k2, trainLen, trainLen, True)
folds = []
s = svm.SVM(C=C)
s.train(trainData)
# testData = SparseDataSet(testFeatureFile);
testData = demo_utils.get_spectrum_data(testSeqFile, k1, k2, testLen, testLen, True)
results = s.test(testData)
labels = results.getGivenClass()
dvals = results.getDecisionFunction()
folds.append((dvals, labels))
demo_utils.print_results(results)
print "Results Log: "
results.getLog()
from PyML import * import demo_utils import generate_model ## Program starts here. numArgs = len(sys.argv); print numArgs if numArgs < 4: print "USAGE: python createTestSpectrumData.py <test seq file> <lenTest> <k1> <k2> <testDataFile>" exit(1); testSeqFile = sys.argv[1]; testLen = int(sys.argv[2]); #equal pos/neg len k1 = int(sys.argv[3]); k2 = int(sys.argv[4]); testDatFile = sys.argv[5]; testData = generate_model.get_spectrum_data(testSeqFile, k1, k2, testLen, testLen, True); testData.save(testDatFile); if __name__ == '__main__': import sys combinedFile = sys.argv[1] print "Combined File: ", combinedFile; prefix = sys.argv[2] confFile = sys.argv[3] CreateConf(combinedFile, prefix, confFile)
confMap = readConfFile(configFile); return confMap; if __name__ == '__main__': import sys confMap = ReadConfFile(sys.argv[1]); dataFile = confMap["data"]["dataFile"] k1 = int(confMap["kspectrum"]["k1"]) k2 = int(confMap["kspectrum"]["k2"]) C = float(confMap["kspectrum"]["C"]) posLen = int(confMap["data"]["posLen"]) negLen = int(confMap["data"]["negLen"]) numFolds = int(confMap["kspectrum"]["folds"]); featureFile = confMap["output"]["featureFile"]; resultFile = confMap["output"]["resultsFile"] modelFile = confMap["output"]["modelFile"] buildModel = False; trainData = generate_model.get_spectrum_data(dataFile, k1, k2, posLen, negLen, normalize=True, repeatCount=False); if buildModel: results, m = BuildModel(trainData, modelFile, numFolds); print_results(results, resultFile, k1, k2) find_features(trainData, featureFile, m.classifier.C); else: results = CrossValidate(trainData, numFolds) print_results(results, resultFile, k1, k2) find_features(trainData, featureFile);
bestK1 = 0;
bestK2 = 0;
Cs = [ 10**x for x in xrange( -10, 5 ) ]
K1 = [8, 9, 10, 11, 12, 13]
K2 = [8, 9, 10, 11, 12, 13]
Cs = [ 10**x for x in xrange( -10, 1 ) ]
K1 = [8]
K2 = [9]
for k1 in K1:
for k2 in [x for x in K2 if x >= k1]:
for C in Cs:
print "**** Train/Test with K1: " + str(k1) + ", k2: " + str(k2) +", C: " + str(C);
trainData = generate_model.get_spectrum_data(trainSeqFile, k1, k2, trainPosLen, trainNegLen, normalize=True, repeatCount=True);
s = svm.SVM(C=C);
s.train(trainData)
results = s.stratifiedCV(trainData, numFolds=numFolds, seen=1);
labels = results.getGivenClass();
roc = results.getROC();
roc50 = results.getROCn();
successRate = results.getSuccessRate();
print "ROC: " + str(roc) + ", ROC50: " + str(roc50) + ", Success Rate: " + str(successRate);
if roc > bestROC:
# Delete the old data file.
os.remove(dataFile) if os.path.exists(dataFile) else None
# Save the training data for K1, K2 combination.
trainData.save(dataFile);
