# ------ plot testing ------
y = np.concatenate([
training_scaler_Y.inverse_transform(trainingY),
training_scaler_Y.inverse_transform(testY)
])
y_true = y.reshape(y.shape[0], )
yhats_training_mean_conv = yhats_training_mean_conv.reshape(
yhats_training_mean_conv.shape[0], )
yhats_training_std = yhats_training_std.reshape(yhats_training_std.shape[0], )
yhats_training_sem = yhats_training_sem.reshape(yhats_training_sem.shape[0], )
yhats_test_mean_conv = yhats_test_mean_conv.reshape(
yhats_test_mean_conv.shape[0], )
yhats_test_std = yhats_test_std.reshape(yhats_test_std.shape[0], )
yhats_test_sem = yhats_test_sem.reshape(yhats_test_sem.shape[0], )
y_yhat_plot(filepath=os.path.join(
res_dir, 'new_lstm_phase1_base_scatter_freq' + str(freq) + '.pdf'),
y_true=y_true,
training_yhat=yhats_training_mean_conv,
training_yhat_err=yhats_training_std,
test_yhat=yhats_test_mean_conv,
test_yhat_err=yhats_test_std,
plot_title='Cross-validation prediction',
ylabel='PCL',
xlabel='Subjects',
plot_type='scatter',
bar_width=0.25)
# ------ true test realm ------
yhats_trainingX_pred = yhats_trainingX_pred.reshape(
yhats_trainingX_pred.shape[0], )
yhats_trainingX_std = yhats_trainingX_std.reshape(
yhats_trainingX_std.shape[0], )
yhats_trainingX_sem = yhats_trainingX_sem.reshape(
yhats_trainingX_sem.shape[0], )
yhats_testX_pred = yhats_testX_pred.reshape(yhats_testX_pred.shape[0], )
yhats_testX_std = yhats_testX_std.reshape(yhats_testX_std.shape[0], )
yhats_testX_sem = yhats_testX_sem.reshape(yhats_testX_sem.shape[0], )
y_yhat_plot(filepath=os.path.join(res_dir,
'new_freq1_cv_plot_scatter_test.pdf'),
y_true=y_true,
training_yhat=yhats_trainingX_pred,
training_yhat_err=yhats_trainingX_std,
test_yhat=yhats_testX_pred,
test_yhat_err=yhats_testX_std,
plot_title='Cross-validation prediction',
ylabel='PCL',
xlabel='Subjects',
plot_type='scatter',
bar_width=0.25)
# ------ training/test subject split for all the frquencies ------
# theta
training, test, scaler_X, scaler_Y = training_test_spliter(
data=raw,
man_split=True,
man_split_colname='subject',
man_split_testset_value=['PN14', 'PN27', 'PP13'],
min_max_scaling=True,
scale_column_as_y=['PCL'],
training_y_hat, test_y_hat = inverse_norm_y(training_y=training_y_hat,
test_y=test_y_hat,
scaler=scaler_Y)
# ---- plot
# -- data for plotting
# NOTE: run once!
y_plot = np.concatenate([lstm_training_y, lstm_test_y])
y_plot = scaler_Y.inverse_transform(y_plot.reshape(y_plot.shape[0], 1))
y_plot = y_plot.reshape(y_plot.shape[0], )
# -- plotting
y_yhat_plot(filepath=os.path.join(res_dir, freq + '_simple.performance.pdf'),
y_true=y_plot,
training_yhat=training_y_hat,
test_yhat=test_y_hat,
plot_title='Simple LSTM model prediction plot',
ylabel='PCL',
xlabel='Subjects',
plot_style='classic')
# for enc-dec model
training_y_hat = training_y_hat.reshape(training_y_hat.shape[0], )
test_y_hat = test_y_hat.reshape(test_y_hat.shape[0], )
y_yhat_plot(filepath=os.path.join(res_dir, freq + '_enc-dec.performance.pdf'),
y_true=y_plot,
training_yhat=training_y_hat,
test_yhat=test_y_hat,
plot_title='Encoder-Decoder LSTM model prediction plot',
ylabel='PCL',
xlabel='Subjects',
plot_type='bar',