def plot_feature_importances(clf,
feature_names,
clf_name='',
top_n=10,
figsize=(8, 8),
T=None,
save_fig=True):
'''
clf: fitted classifier
feature_names = train_x.columns
train_x is pandas dataFrame
'''
title = "Feature Importances of " + clf_name
feat_imp = pd.DataFrame({'importance': clf.feature_importances_})
feat_imp['feature'] = feature_names
feat_imp.sort_values(by='importance', ascending=False, inplace=True)
feat_imp = feat_imp.iloc[:top_n]
feat_imp.sort_values(by='importance', inplace=True)
feat_imp = feat_imp.set_index('feature', drop=True)
feat_imp.plot.barh(title=title, figsize=figsize)
plt.xlabel('Feature Importance Score')
if save_fig == True:
if T == None:
utils.savePlots('dt', plt, name=clf_name)
else:
utils.savePlots('dt', plt, name=clf_name + str(T))
else:
plt.close()
print("not saving the plot of " + clf_name)
#plt.show()
return feat_imp
def getPerformanceMeasures(dataDict, prefix="", T=5, kernel='linear'):
for key in dataDict:
print(">> Dataset: " + key.upper())
[train_x, train_y, test_x, test_y] = dataDict[key]
#Train AdaBoost with SVM
y_ada_svm = buildAdaSVM(train_x, train_y, test_x, T=T, kernel=kernel)
print(">> AdaSVM: \n")
print("Accuracy: " + str(round(metrics.accuracy_score(test_y, y_ada_svm),3)) +
", Precision: " + str(round(metrics.precision_score(test_y, y_ada_svm, average='macro'),3)) +
", Recall: " + str(round(metrics.recall_score(test_y, y_ada_svm, average='macro'),3))
)
utils.plotCM(metrics.confusion_matrix(test_y, y_ada_svm), title= "Adaboost with SVM: " + key.capitalize())
utils.savePlots(modelName, plt, prefix + "ada_" + key, "png")
#Train plain SVM
y_svm = buildSVM(train_x, train_y, test_x, kernel=kernel)
print(">> SVM: \n")
print("Accuracy: " + str(round(metrics.accuracy_score(test_y, y_svm),3)) +
", Precision: " + str(round(metrics.precision_score(test_y, y_svm, average='macro'),3)) +
", Recall: " + str(round(metrics.recall_score(test_y, y_svm, average='macro'),3))
)
utils.plotCM(metrics.confusion_matrix(test_y, y_svm), title= "SVM: " + key.capitalize())
utils.savePlots(modelName, plt, prefix + "svm_" + key, "png")
def plotKernelComparison(dataDict, prefix=""):
plt.clf()
for key in dataDict:
[train_x, train_y, test_x, test_y] = dataDict[key]
accuracies = compareKernels(train_x, train_y, test_x, test_y,)
plt.plot(kernels, accuracies, color=colors[key], label=key.capitalize())
plt.xlabel('SVM Kernel')
plt.ylabel('Accuracy Score')
plt.title('Comparison of SVM kernels in AdaBoost')
plt.legend()
utils.savePlots(modelName, plt, prefix + 'kernel_comparison', "png")
def plotAlphaComparison(dataDict):
plt.clf()
for key in dataDict:
print(">> Dataset: " + key.upper())
[train_x, train_y, test_x, test_y] = dataDict[key]
accuracies = compareAlpha(train_x, train_y, test_x, test_y)
plt.plot(accuracies, color=colors[key], label=key.capitalize())
plt.xlabel('Value of alpha (a)')
plt.ylabel('Accuracy Score')
plt.title('MLP')
plt.legend()
utils.savePlots(modelName, plt, 'alpha_comparison', "png")
def plotTComparison(dataDict, prefix=""):
plt.clf()
for kernel in kernels:
print(kernel.upper())
for key in dataDict:
print(">> Dataset: " + key.upper())
[train_x, train_y, test_x, test_y] = dataDict[key]
accuracies = compareTs(train_x, train_y, test_x, test_y, kernel)
#taskAccuracies[key] = tAccuracies
plt.plot(T, accuracies, color=colors[key], label=key.capitalize())
plt.xlabel('Number of estimators (T)')
plt.ylabel('Accuracy Score')
plt.title('AdaBoost with SVM')
plt.legend()
utils.savePlots(modelName, plt, prefix + kernel + '_t_comparison', "png")
plot_colors = ['red', 'blue', 'green']
line_labels = ['beach', 'finger', 'rest']
count = 0
for data in total_data:
train_x, train_y, test_x, test_y = data_separation(data)
#iteration_num = [1, 5, 10, 50, 100, 500, 1000]
#test_err_list, polar_y_ada = error_list_ada(iteration_num, accuracy=True)
clf_tree = tree.DecisionTreeClassifier()
y_pred = clf_tree.fit(train_x, train_y).predict(test_x)
svm_clf = SVC(gamma='auto', kernel='poly', degree=3, coef0=0.015625)
y_svm = svm_clf.fit(train_x, train_y).predict(test_x)
graph = confusion_matrix(test_y, y_svm)
utils.plotCM(graph, title="SVM - Polynomial: " + line_labels[count])
utils.savePlots("dt", plt, "CM_SVM_ARMP_" + line_labels[count], "png")
#plt.plot(iteration_num, test_err_list, color = plot_colors[count], label=line_labels[count])
count += 1
plt.legend()
plt.xlabel('Number of estimators (T)')
plt.ylabel('Accuracy Score')
plt.title('AdaBoost with Decision Stump')
#graph = confusion_matrix(test_y, y_gnb)
'''
svm_clf = SVC(gamma='auto')
y_svm = svm_clf.fit(train_x,train_y).predict(test_x)
#gmm = GaussianMixture(n_components=20)
#y_gmm = gmm.fit(train_x,train_y).predict(test_x)