print('acc:', clf.score(np.array(testX), np.array(testY)))
print("\n====== Random Forest ======")
for c in [10, 30, 50, 70, 90, 100, 120]:
clf = RandomForestClassifier(n_estimators=c)
clf.fit(trainX, trainY)
proY = clf.predict_proba(testX)
print(">>> # tree: ", c)
print('other measure: ', scoreother_measured(proY, testY))
print("acc: ", clf.score(np.array(testX), np.array(testY)))
print("\n====== SVM ======")
# for c in [0.8, 1.0]:
# clf = SVC(C =c,probability = True)
# clf.fit(trainX, trainY)
# clf.probability = True
# proY = clf.predict_proba(testX)
# print('add other_measured acc', scoreother_measured(proY, testY))
# print('SVM acc when penalty C = ', c, clf.score(np.array(testX), np.array(testY)))
print("----- kernel -----")
print("----this will take more time, please be patient....")
for c in ['linear', 'poly', 'rbf']:
clf = SVC(kernel=c, probability=True) #, probability = True)
clf.fit(trainX, trainY)
clf.probability = True
# proY = clf.predict_proba(testX)
proY = clf.predict_proba(testX)
print(">>> kernel:", c)
print('other measure: ', scoreother_measured(proY, testY))
print('acc:', clf.score(np.array(testX), np.array(testY)))
ax.set_xticklabels(data.columns)
ax.set_yticklabels(data.columns)
fig.savefig('correlation.png')
#fig.show()
# Plot ROC Curve
plt.clf()
plt.figure(figsize=(8, 6))
d_tree.probability = True
probas = d_tree.predict_proba(X_test_d)
fpr, tpr, thresholds = met.roc_curve(y_test_d, probas[:, 1])
roc_auc = met.auc(fpr, tpr)
plt.plot(fpr, tpr, label='%s ROC (area = %0.2f)' % ('Decision Tree', roc_auc))
naivebayes.probability = True
probas = naivebayes.predict_proba(X_test_n)
fpr, tpr, thresholds = met.roc_curve(y_test_n, probas[:, 1])
roc_auc = met.auc(fpr, tpr)
plt.plot(fpr, tpr, label='%s ROC (area = %0.2f)' % ('Naive Bayes', roc_auc))
probas = rbf_kernel_clf.predict_proba(X_test_s)
fpr, tpr, thresholds = met.roc_curve(y_test_s, probas[:, 1])
roc_auc = met.auc(fpr, tpr)
plt.plot(fpr, tpr, label='%s ROC (area = %0.2f)' % ('SVM', roc_auc))
mlp.probability = True
probas = mlp.predict_proba(X_test_a)
fpr, tpr, thresholds = met.roc_curve(y_test_a, probas[:, 1])
roc_auc = met.auc(fpr, tpr)
plt.plot(fpr, tpr, label='%s ROC (area = %0.2f)' % ('Neural Network', roc_auc))
