bow.patch_size=(20,20)
bow.learning_rate=0.001
bow.n_components=512
bow.n_iter=100
bow.sample_num = 1000
bow.fit(x_train)
svm = SVC(kernel='linear', probability = True, random_state=42)
svm.C = 1000
#lr = LogisticRegression()
#lr.C = 100
best = Pipeline([('bow', bow),('svm',svm)])
best.fit(x_train, y_train)
print "*********************Save*******************************"
joblib.dump(best, "classifier_rbm.pkl", compress=3)
print "*********************Test*******************************"
y_test_pre = best.predict(x_test)
cm = confusion_matrix(y_test, y_test_pre)
from map_confusion import plot_conf
plot_conf(cm, range(le.classes_.size), 'RSDataset.png')
from sklearn.metrics import classification_report
with file('report_rbm.txt', 'w') as f:
report = classification_report(y_test, y_test_pre, target_names = le.classes_)
f.writelines(report)
show(bow.rbm.components_,(20,20))
if spm.kernel_hi:
svm = SVC(kernel='precomputed', probability = True,random_state=42)
else:
svm = SVC(kernel='linear', probability = True,random_state=42)
clf = Pipeline([('spm', spm),('svm',svm)])
best = Pipeline([('spm', spm),('svm',svm)])
best.fit(x_train, y_train)
print "*********************Save*******************************"
if method != 'sift':
joblib.dump(best, "classifier_spm_%s.pkl"%method, compress=3)
print "*********************Test*******************************"
spm = best.named_steps['spm']
y_test_pre = best.predict(x_test)
from sklearn.metrics import accuracy_score
print "Accuracy:", accuracy_score(y_test, y_test_pre)
cm = confusion_matrix(y_test, y_test_pre)
np.save('RS_results/RSDataset_%s_%s_%s.npy'%(method, spm.clusters, spm.img_size), cm)
from map_confusion import plot_conf
plot_conf(conf_arr=cm, label_list=range(le.classes_.size),
norm=False, save_name='RS_results/RSDataset_%s_%s_%s.png'%(method, spm.clusters, spm.img_size))
from sklearn.metrics import classification_report
with file('RS_results/report_spm_%s_%s_%s.txt'%(method, spm.clusters, spm.img_size), 'w') as f:
report = classification_report(y_test, y_test_pre, target_names = le.classes_)
f.writelines(report)
#show(bow.rbm.components_,(20,20))
X.append(item)
Y = le.transform(Y)
X = np.vstack(X)
all_x.append(X)
all_y.append(Y)
all_x = np.vstack(all_x)
all_y = np.hstack(all_y)
cv = StratifiedShuffleSplit(y=all_y, n_iter=1, test_size=0.4)
for train, test in cv:
train_x, train_y, test_x, test_y = all_x[train], all_y[train], all_x[test], all_y[test]
clf = SVC(C=1, kernel='linear', probability = True, random_state=42)
print "Training..."
clf.fit(train_x, train_y)
y_test_pre = clf.predict(test_x)
print "Confusion matrix..."
from sklearn.metrics import accuracy_score
from sklearn.metrics import confusion_matrix
print "Accuracy:", accuracy_score(test_y, y_test_pre)
cm = confusion_matrix(test_y, y_test_pre)
np.save('RS_results/RSDataset_%s_%s_%s.npy'%('sift', 1000, 600), cm)
from map_confusion import plot_conf
plot_conf(conf_arr=cm, label_list=le.classes_.tolist(),
norm=False, save_name='RS_results/RSDataset_%s_%s_%s.png'%('sift', 1000, 600))
from sklearn.metrics import classification_report
with file('RS_results/report_spm_%s_%s_%s.txt'%('sift', 1000, 600), 'w') as f:
report = classification_report(test_y, y_test_pre, target_names = le.classes_)
f.writelines(report)
print "*********************Train******************************"
best = gridCV.best_estimator_
best.fit(x_train, y_train)
print "*********************Save*******************************"
joblib.dump(best, "classifier.pkl", compress=3)
joblib.dump(gridCV, "grid_cv.pkl", compress=3)
# best = joblib.load(prj_name + "/classifier_svc.pkl")
print "*********************Test*******************************"
y_test_pre = best.predict(x_test)
cm = confusion_matrix(y_test, y_test_pre)
from map_confusion import plot_conf
plot_conf(cm, range(le.classes_.size))
#print "confusion matrix..."
#print cm
from sklearn.metrics import classification_report
with file('report.txt', 'w') as f:
report = classification_report(y_test, y_test_pre, target_names = le.classes_)
f.writelines(report)
"""
print "*********************ROC********************************"
from sklearn.metrics import roc_curve, auc
from sklearn.preprocessing import label_binarize
print best
y_score = best.predict_proba(x_test)
