def predict_proba(self, X, input_checks=True, **kwargs):
"""
Find probability estimates for each class for all cases in X.
Parameters
----------
X : a nested pd.Dataframe, or (if input_checks=False) array-like of
shape = (n_instances, series_length, n_dimensions)
The training input samples. If a 2D array-like is passed,
n_dimensions is assumed to be 1.
input_checks: boolean
whether to check the X parameter
Returns
-------
output : array of shape = [n_instances, n_classes] of probabilities
"""
check_is_fitted(self)
X = check_and_clean_data(X, input_checks=input_checks)
probs = self.model.predict(X, **kwargs)
# check if binary classification
if probs.shape[1] == 1:
# first column is probability of class 0 and second is of class 1
probs = np.hstack([1 - probs, probs])
return probs
def predict_proba(self, X, input_checks=True, **kwargs):
"""
Find probability estimates for each class for all cases in X.
Parameters
----------
X : a nested pd.Dataframe, or (if input_checks=False) array-like of
shape = (n_instances, series_length, n_dimensions)
The training input samples. If a 2D array-like is passed,
n_dimensions is assumed to be 1.
input_checks: boolean
whether to check the X parameter
Returns
-------
output : array of shape = [n_instances, n_classes] of probabilities
"""
check_is_fitted(self)
X = check_and_clean_data(X, input_checks=input_checks)
# transform and predict prodba on the ridge classifier.
X_transformed = self.transform_to_feature_space(X)
y_pred = self.model.predict(X_transformed)
# self.reshape_prediction will give us PREDICTIONS,
# not DISTRIBUTIONS (even if only one-hot)
# Computing first 2 lines of that but not the last here
# reshape so the first axis has the number of instances
new_y_pred = y_pred.reshape(X.shape[0], X.shape[1], y_pred.shape[-1])
# average the predictions of instances
return np.average(new_y_pred, axis=1)
def predict(self, X, input_checks=True, **kwargs):
"""
Find regression estimate for all cases in X.
Parameters
----------
X : a nested pd.Dataframe, or (if input_checks=False) array-like of
shape = (n_instances, series_length, n_dimensions)
The training input samples. If a 2D array-like is passed,
n_dimensions is assumed to be 1.
input_checks: boolean
whether to check the X parameter
Returns
-------
predictions : 1d numpy array
array of predictions of each instance
"""
check_is_fitted(self)
X = check_and_clean_data(X, input_checks=input_checks)
y_pred = self.model.predict(X, **kwargs)
if y_pred.ndim == 1:
y_pred.ravel()
return y_pred
def predict(self, X, input_checks=True, **kwargs):
"""
Find regression estimate for all cases in X.
Parameters
----------
X : a nested pd.Dataframe, or (if input_checks=False) array-like of
shape = (n_instances, series_length, n_dimensions)
The training input samples. If a 2D array-like is passed,
n_dimensions is assumed to be 1.
input_checks: boolean
whether to check the X parameter
Returns
-------
predictions : 1d numpy array
array of predictions of each instance
"""
check_is_fitted(self)
X = check_and_clean_data(X, input_checks=input_checks)
X, _, tot_increase_num = self.pre_processing(X)
preds = self.model.predict(X, batch_size=self.batch_size)
y_predicted = []
test_num_batch = int(X.shape[0] / tot_increase_num)
# TODO: could fix this to be an array literal.
for i in range(test_num_batch):
y_predicted.append(
np.average(
preds[i * tot_increase_num:((i + 1) * tot_increase_num) -
1],
axis=0,
))
y_pred = np.array(y_predicted)
if y_pred.ndim == 1:
y_pred.ravel()
return y_pred
def predict_proba(self, X, input_checks=True, **kwargs):
"""
Find probability estimates for each class for all cases in X.
Parameters
----------
X : a nested pd.Dataframe, or (if input_checks=False) array-like of
shape = (n_instances, series_length, n_dimensions)
The training input samples. If a 2D array-like is passed,
n_dimensions is assumed to be 1.
input_checks: boolean
whether to check the X parameter
Returns
-------
output : array of shape = [n_instances, n_classes] of probabilities
"""
check_is_fitted(self)
X = check_and_clean_data(X, input_checks=input_checks)
X, _, tot_increase_num = self.pre_processing(X)
preds = self.model.predict(X, batch_size=self.batch_size)
test_num_batch = int(X.shape[0] / tot_increase_num)
y_predicted = [
np.average(
preds[i * tot_increase_num:((i + 1) * tot_increase_num) - 1],
axis=0)
for i in range(test_num_batch)
]
y_pred = np.array(y_predicted)
keras.backend.clear_session()
return y_pred
def predict_proba(self, X, input_checks=True, **kwargs):
"""
Find probability estimates for each class for all cases in X.
Parameters
----------
X : a nested pd.Dataframe, or (if input_checks=False) array-like of
shape = (n_instances, series_length, n_dimensions)
The training input samples. If a 2D array-like is passed,
n_dimensions is assumed to be 1.
input_checks: boolean
whether to check the X parameter
Returns
-------
output : array of shape = [n_instances, n_classes] of probabilities
"""
check_is_fitted(self)
X = check_and_clean_data(X, input_checks=input_checks)
ori_len = X.shape[1] # original_length of time series
# restrict slice ratio when data lenght is too large
current_slice_ratio = self.slice_ratio
if ori_len > 500:
current_slice_ratio = (
self.slice_ratio if self.slice_ratio > 0.98 else 0.98
)
increase_num = (
ori_len - int(ori_len * current_slice_ratio) + 1
) # this can be used as the bath size
# will need to slice at some poin
x_test, _ = self.slice_data(X, None, current_slice_ratio)
length_train = x_test.shape[1] # length after slicing.
current_window_size = (
int(length_train * self.window_size)
if self.window_size < 1
else int(self.window_size)
)
ds_num_max = length_train / (
self.best_pool_factor * current_window_size
)
current_ds_num = int(min(self.ds_num, ds_num_max))
# need to batch and downsample the test data.
ma_test, ma_lengths = self.movingavrg(
x_test, self.ma_base, self.ma_step, self.ma_num
)
ds_test, ds_lengths = self.downsample(
x_test, self.ds_base, self.ds_step, current_ds_num
)
test_set_x = x_test
# concatenate directly
data_lengths = [length_train]
# downsample part:
if ds_lengths != []:
data_lengths += ds_lengths
test_set_x = np.concatenate([test_set_x, ds_test], axis=1)
# moving average part
if ma_lengths != []:
data_lengths += ma_lengths
test_set_x = np.concatenate([test_set_x, ma_test], axis=1)
test_num = x_test.shape[0]
test_num_batch = int(test_num / increase_num)
# get the true predictions of the test set
y_predicted = []
for i in range(test_num_batch):
x = test_set_x[i * (increase_num): (i + 1) * (increase_num)]
preds = self.model.predict_on_batch(
self.split_input_for_model(x, self.input_shapes)
)
y_predicted.append(np.average(preds, axis=0))
y_pred = np.array(y_predicted)
return y_pred
