def knn_shapelet_experiment(knnClassifier, X_train, y_train, X_test, y_test, num_resample=1, verbose=True):
minShapeletLength = 30
maxShapeletLength = int(X_train.shape[1] * 0.7)
numShapelet = int(X_train.shape[1] * 0.5)
lengthIncrement = 20
positionIncrement = 20
sum_accuracy = 0
for i in range(num_resample):
callSTS = ShapeletTransformSimplicity(X_train, y_train,
minShapeletLength, maxShapeletLength,
lengthIncrement, positionIncrement)
bestKShapelet = callSTS.train(numShapelet)
print("bestKShapelet: ", np.shape(bestKShapelet))
X_train_ = callSTS.transform(bestKShapelet, X_train)
X_test_ = callSTS.transform(bestKShapelet, X_test)
X_train_ = np.array(X_train_)
X_test_ = np.array(X_test_)
knnClassifier.fit(X_train_, y_train)
predList = knnClassifier.predict(X_test_)
iaccuracy = validation.cal_accuracy(y_test, predList)
sum_accuracy += iaccuracy
if verbose:
print(iaccuracy)
X_train, y_train, X_test, y_test = data_parser.resample_data(X_train, y_train, X_test, y_test)
return sum_accuracy / num_resample
def benchmark(clf, distf, distparams=None):
t0 = time()
clf.set_distfunc(distfunc=distf, distfunc_params=distparams)
y_pred = clf.predict(X_test)
duration = time() - t0
acc = validation.cal_accuracy(y_pred, y_test)
return acc, duration
def knn_experiment(knnClassifier, X_train, y_train, X_test, y_test, num_resample=1, verbose=True):
sum_accuracy = 0
for i in range(num_resample):
knnClassifier.fit(X_train, y_train)
predList = knnClassifier.predict(X_test)
iaccuracy = validation.cal_accuracy(y_test, predList)
if verbose:
print(iaccuracy)
sum_accuracy += iaccuracy
X_train, y_train, X_test, y_test = data_parser.resample_data(X_train, y_train, X_test, y_test)
return sum_accuracy / num_resample
print("file_test", file_test)
X_train, y_train = data_parser.load_list_data(file_train)
X_test, y_test = data_parser.load_list_data(file_test)
print('test class', np.unique(y_train))
print('train class', np.unique(y_test))
# print(X_train[0])
# print(X_test[0])
# z_normalize
for id, iTrain in enumerate(X_train):
X_train[id] = (iTrain - np.mean(iTrain)) / (np.sqrt(np.var(iTrain)) +
1e-9)
for id, iTest in enumerate(X_test):
X_test[id] = (iTest - np.mean(iTest)) / (np.sqrt(np.var(iTest)) + 1e-9)
# print(X_train[0])
# print(X_test[0])
n_neighbors = 1
n_jobs = 10
knn_clf = KNeighborsClassifierVariousDim(n_neighbors=n_neighbors,
distfunc=DISTFunc.dist_basic_dtw)
knn_clf.fit(X_train, y_train)
predList = knn_clf.predict(X_test)
acc = validation.cal_accuracy(predList, y_test)
print(acc)
n_neighbors = 1
n_jobs = 10
k_fold = 10
distfunc = distance.dist_euclidean
# set up knn classifier
knn_clf = KNeighborsClassifier(n_neighbors=n_neighbors,
n_jobs=n_jobs,
distfunc=distfunc)
result = [['dataset name', 'accuracy']]
for name in dataset_:
file_train = os.path.join(UCR_DATA_ROOT, name, name + '_TRAIN')
file_test = os.path.join(UCR_DATA_ROOT, name, name + '_TEST')
X_train, y_train = data_parser.load_ucr(file_train)
X_test, y_test = data_parser.load_ucr(file_test)
acc_sum = 0
for i in range(k_fold):
knn_clf.fit(X_train, y_train)
y_pred = knn_clf.predict(X_test)
acc_sum += validation.cal_accuracy(y_test, y_pred)
X_train, y_train, X_test, y_test = data_parser.resample_data(
X_train, y_train, X_test, y_test)
acc = acc_sum / k_fold
result.append((name, acc))
print(name, acc)
file_out = os.path.join(OUT_ROOT, "%dnn_%dfold_ed.md" % (n_neighbors, k_fold))
table2markdown(file_out, result, description=__doc__)