def runRegressionModelTest(featureSet, valueVector, model):
output = ''
clf = 0
if model == 1:
print "\nLINEAR REGRESSION\n"
clf = linear_regression_fit(featureSet, valueVector)
elif model == 2:
print "\nSVR\n"
clf = SVR_fit(featureSet, valueVector)
elif model == 4:
print "\nSTOCHASTIC\n"
clf = SGD_r_fit(featureSet, valueVector)
joblib.dump(clf, 'sgd.pkl')
elif model == 5:
print "\nNEIGHBOURS\n"
clf = neighbours_fit(featureSet, valueVector)
elif model == 6:
print "\nLOGISTIC\n"
clf = log_regression_fit(featureSet, valueVector)
elif model == 7:
print "\nBAYESIANRIDGE\n"
clf = bayesian_ridge_fit(featureSet, valueVector)
elif model == 8:
print "\nRIDGE\n"
clf = ridge_fit(featureSet, valueVector)
elif model == 9:
print "\nELASTIC NET\n"
clf = elastic_fit(featureSet, valueVector)
elif model == 10:
print "\nLASSO\n"
clf = lasso_fit(featureSet, valueVector)
else:
print 'Invalid choice\n'
return clf
def runRegressionModelTest(featureSet, valueVector, X_test, y_test, model):
output = ''
score = 0
clf = 0
if model == 1:
output += "\nLINEAR REGRESSION\n"
clf = linear_regression_fit(featureSet, valueVector)
elif model == 2:
output += "\nSVR\n"
clf = SVR_fit(featureSet, valueVector)
elif model == 3:
output += "\nEXTREME LEARNING MACHINE\n"
clf = elm.ELMRegressor()
clf.fit(featureSet, valueVector)
joblib.dump(clf, 'elm.pkl')
elif model == 4:
output += "\nSTOCHASTIC\n"
clf = SGD_r_fit(featureSet, valueVector)
joblib.dump(clf, 'sgd.pkl')
elif model == 5:
output += "\nNEIGHBOURS\n"
clf = neighbours_fit(featureSet, valueVector)
elif model == 6:
output += "\nLOGISTIC\n"
clf = log_regression_fit(featureSet, valueVector)
elif model == 7:
output += "\nBAYESIANRIDGE\n"
clf = bayesian_ridge_fit(featureSet, valueVector)
else:
output += 'Invalid choice\n'
score = mean_squared_error(y_test, clf.predict(X_test))
score2 = r2_score(y_test, clf.predict(X_test))
cv = cross_validation.ShuffleSplit(featureSet.shape[0],
n_iter=50,
test_size=0.25,
random_state=0)
a = cross_validation.cross_val_score(clf, featureSet, valueVector, cv=cv)
a = a[a > 0]
output += 'Cross V score: :' + ' '.join("%10.3f" % x for x in a) + '\n'
output += ('Mean Score: %.3f\n' % np.mean(a))
output += ('Mean Squared Error: %.3f\n' % score)
output += ('R^2: %.3f\n' % score2)
return output
def runRegressionModelTest(featureSet, valueVector, X_test, y_test, model):
output = ''
score = 0
clf = 0
if model == 1:
output += "\nLINEAR REGRESSION\n"
clf = linear_regression_fit(featureSet, valueVector)
elif model == 2:
output += "\nSVR\n"
clf = SVR_fit(featureSet, valueVector)
elif model == 3:
output += "\nEXTREME LEARNING MACHINE\n"
clf = elm.ELMRegressor()
clf.fit(featureSet, valueVector)
joblib.dump(clf, 'elm.pkl')
elif model == 4:
output += "\nSTOCHASTIC\n"
clf = SGD_r_fit(featureSet, valueVector)
joblib.dump(clf, 'sgd.pkl')
elif model == 5:
output += "\nNEIGHBOURS\n"
clf = neighbours_fit(featureSet, valueVector)
elif model == 6:
output += "\nLOGISTIC\n"
clf = log_regression_fit(featureSet, valueVector)
elif model == 7:
output += "\nBAYESIANRIDGE\n"
clf = bayesian_ridge_fit(featureSet, valueVector)
else :
output += 'Invalid choice\n'
score = mean_squared_error(y_test, clf.predict(X_test))
score2 = r2_score(y_test, clf.predict(X_test))
cv = cross_validation.ShuffleSplit(featureSet.shape[0], n_iter=50,test_size=0.25,random_state=0)
a = cross_validation.cross_val_score(clf, featureSet, valueVector, cv=cv)
a = a[a > 0]
output += 'Cross V score: :' + ' '.join("%10.3f" % x for x in a) + '\n'
output += ('Mean Score: %.3f\n' % np.mean(a))
output += ('Mean Squared Error: %.3f\n' % score)
output += ('R^2: %.3f\n' % score2)
return output
def runClassificationTest(X, y, Xt, yt, model, labs):
output = ''
clf = 0
if model == 1:
output += "\nSVC\n"
clf = svc_fit(X, y)
elif model == 2:
output += '\nLinearSVC\n'
clf = linear_svc_fit(X, y)
elif model == 3:
output += '\nStochasticGradientDescent\n'
clf = SGD_c_fit(X, y)
elif model == 4:
output += '\nKNearestNeighbours\n'
clf = nearest_fit(X, y)
elif model == 5:
output += '\nRandomForest\n'
clf = random_forest_fit(X, y)
elif model == 6:
output += '\nLogistic\n'
clf = log_regression_fit(X, y)
accuracy = accuracy_score(yt, clf.predict(Xt))
f1 = f1_score(yt, clf.predict(Xt), labels=labs)
cv = cross_validation.ShuffleSplit(X.shape[0], n_iter=50,test_size=0.3,random_state=0)
a = cross_validation.cross_val_score(clf, X, y, cv=cv)
a = a[a > 0]
output += 'Cross V score: :' + ' '.join("%10.3f" % x for x in a) + '\n'
output += "\n\nAccuracy " + str(accuracy)
output += "\nF1 Score " + str(f1)
return clf, output;
def runClassificationTest(X, y, Xt, yt, model, labs):
output = ''
clf = 0
if model == 1:
output += "\nSVC\n"
clf = svc_fit(X, y)
elif model == 2:
output += '\nLinearSVC\n'
clf = linear_svc_fit(X, y)
elif model == 3:
output += '\nStochasticGradientDescent\n'
clf = SGD_c_fit(X, y)
elif model == 4:
output += '\nKNearestNeighbours\n'
clf = nearest_fit(X, y)
elif model == 5:
output += '\nRandomForest\n'
clf = random_forest_fit(X, y)
elif model == 6:
output += '\nLogistic\n'
clf = log_regression_fit(X, y)
accuracy = accuracy_score(yt, clf.predict(Xt))
f1 = f1_score(yt, clf.predict(Xt), labels=labs)
cv = cross_validation.ShuffleSplit(X.shape[0],
n_iter=50,
test_size=0.3,
random_state=0)
a = cross_validation.cross_val_score(clf, X, y, cv=cv)
a = a[a > 0]
output += 'Cross V score: :' + ' '.join("%10.3f" % x for x in a) + '\n'
output += "\n\nAccuracy " + str(accuracy)
output += "\nF1 Score " + str(f1)
return clf, output
def train_classifier(X, y, model, featureset, data_source):
kernel = 'rbf'
parameters = np.zeros([21, 3])
if data_source == 1:
parameters[0] = [10000, 0.001, 10000] # Manual
parameters[1] = [10, 0.01, 0.1] # Unigram
parameters[2] = [10, 0.001, 0.1]
parameters[3] = [10, 0.1, 1]
parameters[4] = [100000, 0.001, 0.1] # Bigram
parameters[5] = [100, 0.1, 10]
parameters[6] = [100, 0.001, 0.1] # Trigram
parameters[7] = [10, 0.1, 10000]
parameters[8] = [1000, 0.001, 1] # Bigram only
parameters[9] = [10, 0.1, 100]
parameters[10] = [0.01, 0.1, 1] # Trigram only
parameters[11] = [0.01, 10, 100]
parameters[12] = [10, 0.0001, 0.001] # Character Ngram
parameters[13] = [10, 1, 10]
parameters[14] = [10, 0.001, 0.01] # Character Skipgram
parameters[15] = [10000, 0.0001, 1]
parameters[16] = [1000000, 10.0, 0.001] # LDA
parameters[17] = [10000, 0.0001, 1] # Word2Vec TFIDF
parameters[18] = [10000000, 0.0001, 10000] # Word2Vec BOC
parameters[19] = [1, 0.01, 10] # Doc2Vec
parameters[20] = [10, 0.001, 0.01] # Google Word2vec TFIDF
elif data_source == 2: # TOY
parameters[0] = [0.1, 0.1, 10] # Manual
parameters[1] = [100, 0.01, 1] # Unigram
parameters[2] = [100, 0.01, 1]
parameters[3] = [10, 1, 10]
parameters[4] = [10, 0.01, 0.1] # Bigram
parameters[5] = [100, 0.1, 1000]
parameters[6] = [10, 0.01, 10] # Trigram
parameters[7] = [10, 0.1, 10000]
parameters[8] = [1000, 0.1, 1] # Bigram only
parameters[9] = [10000, 0.1, 10000]
parameters[10] = [10, 0.1, 1] # Trigram only
parameters[11] = [10, 10, 1000]
parameters[12] = [10, 0.001, 0.1] # Character Ngram
parameters[13] = [100, 0.1, 100]
parameters[14] = [100, 0.001, 0.1] # Character Skipgram
parameters[15] = [10, 1, 10]
parameters[16] = [10000000, 1, 1] # LDA
parameters[17] = [10, 0.01, 0.01] # Word2Vec TFIDF
parameters[18] = [1000, 0.0001, 0.1] # Word2Vec BOC
parameters[19] = [100, 0.0001, 0.01] # Doc2Vec
parameters[20] = [10, 0.001, 1] # Google Word2vec TFIDF
elif data_source == 3: # SLASHDOT
parameters[0] = [1000000, 0.0001, 1000] # Manual
parameters[1] = [1000, 0.1, 1000] # Unigram
parameters[2] = [1000, 0.1, 1]
parameters[3] = [1000, 1, 100]
parameters[4] = [100, 0.001, 1000] # Bigram
parameters[5] = [1, 1, 10000]
parameters[6] = [100, 0.001, 10000] # Trigram
parameters[7] = [1, 0.1, 10000]
parameters[8] = [1000, 0.001, 1] # Bigram only
parameters[9] = [0.1, 1, 100]
parameters[10] = [10, 0.1, 1] # Trigram only
parameters[11] = [100, 1, 1000]
parameters[12] = [1000, 0.001, 0.01] # Character Ngram
parameters[13] = [10, 1, 100]
parameters[14] = [10, 0.001, 0.1] # Character Skipgram
parameters[15] = [10, 1, 10]
parameters[16] = [1000000, 0.01, 10] # LDA
parameters[17] = [10, 0.01, 100] # Word2Vec TFIDF
parameters[18] = [1, 0.001, 0.001] # Word2Vec BOC
parameters[19] = [1000000, 0.0001, 1000] # Doc2Vec
parameters[20] = [10000000, 0.0001, 10000] # Google Word2vec TFIDF
C = parameters[featureset][0]
gamma = parameters[featureset][1]
Lc = parameters[featureset][2]
if model == 1:
print "\nSVC\n"
clf = svc_fit(X, y, kernel=kernel, C=C, gamma=gamma)
elif model == 2:
print '\nLinearSVC\n'
clf = linear_svc_fit(X, y, C=Lc)
elif model == 3:
print '\nStochasticGradientDescent\n'
clf = SGD_c_fit(X, y)
elif model == 4:
print '\nKNearestNeighbours\n'
clf = nearest_fit(X, y)
elif model == 5:
print '\nRandomForest\n'
clf = random_forest_fit(X, y)
elif model == 6:
print '\nLogistic\n'
clf = log_regression_fit(X, y)
return clf