def test_actual_results(params):
"""Test that the actual results are the expected ones."""
knn = KNeighborsClassifier(n_neighbors=1, **params)
proba_actual = knn.fit(X, y).predict_proba(X_test)
pred_actual = knn.predict(X_test)
np.testing.assert_array_equal(proba_actual, y_proba)
np.testing.assert_array_equal(pred_actual, y)
file_train = PATH + str(dataset) + "/" + str(dataset) + "_TRAIN.tsv"
file_test = PATH + str(dataset) + "/" + str(dataset) + "_TEST.tsv"
for i, (dataset, warping_window) in enumerate(zip(dataset_list, warping_window_list)):
print("Dataset: {}".format(dataset))
file_train = PATH + str(dataset) + "/" + str(dataset) + "_TRAIN.tsv"
file_test = PATH + str(dataset) + "/" + str(dataset) + "_TEST.tsv"
train = np.genfromtxt(fname=file_train, delimiter="\t", skip_header=0)
test = np.genfromtxt(fname=file_test, delimiter="\t", skip_header=0)
X_train, y_train = train[:, 1:], train[:, 0]
X_test, y_test = test[:, 1:], test[:, 0]
clf_ed = KNeighborsClassifier(metric='euclidean')
clf_dtw = KNeighborsClassifier(metric='dtw')
clf_dtw_w = KNeighborsClassifier(metric='dtw_sakoechiba',
metric_params={'window_size': warping_window})
transformer = FunctionTransformer(func=lambda x: x.toarray(),
validate=False, check_inverse=False)
knn = KNeighborsClassifier(n_neighbors=1, metric='boss')
# Euclidean Distance
error_ed = 1 - clf_ed.fit(X_train, y_train).score(X_test, y_test)
print('Accuracy ED: ', 1 - error_ed )
print("Error rate with Euclidean Distance: {0:.4f}".format(error_ed))
error_ed_list.append(error_ed)
# file_train = PATH + str(dataset) + "/" + str(dataset) + "_TRAIN"
# file_test = PATH + str(dataset) + "/" + str(dataset) + "_TEST"
# train = np.genfromtxt(fname=file_train, delimiter="\t", skip_header=0)
# test = np.genfromtxt(fname=file_test, delimiter="\t", skip_header=0)
# X_train, y_train = train[:, 1:], train[:, 0]
# X_test, y_test = test[:, 1:], test[:, 0]
direc = '/Users/apple/Desktop/dev/projectlife/data/UCR'
summaries_dir = '/Users/apple/Desktop/dev/projectlife/data/logs'
"""Load the data"""
X_train, X_test, y_train, y_test = load_data(direc, dataset='Projectlife')
clf_ed = KNeighborsClassifier(metric='euclidean')
clf_dtw = KNeighborsClassifier(metric='dtw')
clf_dtw_w = KNeighborsClassifier(
metric='dtw_sakoechiba', metric_params={'window_size': warping_window})
# Euclidean Distance
error_ed = 1 - clf_ed.fit(X_train, y_train).score(X_test, y_test)
print("Error rate with Euclidean Distance: {0:.4f}".format(error_ed))
error_ed_list.append(error_ed)
# Dynamic Time Warping
error_dtw = 1 - clf_dtw.fit(X_train, y_train).score(X_test, y_test)
print("Error rate with Dynamic Time Warping: {0:.4f}".format(error_dtw))
error_dtw_list.append(error_dtw)
print(clf_dtw.predict(X_test))
dtw_sakoechiba, path_sakoechiba = dtw(X_train[i], X_train[j], dist='square', method='sakoechiba',options={'window_size': warping_window}, return_path=True)
DTW_sakoe_train.append(dtw_sakoechiba)
DTW_sakoe_train = np.array(DTW_sakoe_train)
DTW_sakoe_train.resize(y_train.shape[0],int(len(DTW_sakoe_train)/y_train.shape[0]))
DTW_sakoe_test = np.array(DTW_sakoe_test)
DTW_sakoe_test.resize(y_test.shape[0],int(len(DTW_sakoe_test)/y_test.shape[0]))
app.append(1-clf.fit(DTW_Classic_train, y_train).score(DTW_Classic_test,y_test))
app.append(1-gpc.fit(DTW_Classic_train, y_train).score(DTW_Classic_test,y_test))
app.append(1-clf.fit(DTW_sakoe_train, y_train).score(DTW_sakoe_test,y_test))
app.append(1-gpc.fit(DTW_sakoe_train, y_train).score(DTW_sakoe_test,y_test))
print('1NN started')
clf_dtw_w = KNeighborsClassifier(metric='dtw_sakoechiba',metric_params={'window_size': warping_window})
error_dtw_w = 1- clf_dtw_w.fit(X_train, y_train).score(X_test, y_test)
app.append(error_dtw_w)
print('DTW-DTWR started')
from sklearn import svm
from tslearn.preprocessing import TimeSeriesScalerMeanVariance
from pyts.metrics import dtw
from pyts.metrics.dtw import (cost_matrix, accumulated_cost_matrix,
_return_path, _multiscale_region)
test_DTW_DTWR = np.concatenate((DTW_sakoe_test,DTW_Classic_test),axis=1)
train_DTW_DTWR = np.concatenate((DTW_Classic_train,DTW_sakoe_train),axis=1)