def main():
df_train = pd.read_csv('data/train.csv')
df_test = pd.read_csv('data/test.csv')
oof_svc, preds_svc = train_nusvc(df_train, df_test)
ut.plot_results(oof_svc, preds_svc, df_train, 'nusvc.png')
def task2(nmax, random_numbers, steps_count=20):
results = []
# steps_count = 20
step = nmax / steps_count
val_list = range(2, nmax, int(step))
for k in val_list:
tmp = ut.std_dev(k, random_numbers)
results.append(tmp)
# prec_result = [ut.multi_dim_analytical(1)] * len(results)
ut.plot_results(val_list, results, "std_dev Simulation", "log",
"Number of samples", "Integral value")
def task3(nmax, random_numbers, dim):
results = []
# steps_count = 20
val_list = range(1, dim)
prec_results = []
for k in val_list:
print("simul:" + str(ut.multi_dim_approach(nmax, random_numbers, k)))
print("precise:" + str(ut.multi_dim_analytical(k)))
tmp = ut.multi_dim_approach(nmax, random_numbers, k)
results.append(tmp)
prec_results.append(ut.multi_dim_analytical(k))
results = ut.np.asarray(results)
prec_results = ut.np.asarray(prec_results)
diff = results - prec_results
ut.plot_results(val_list, diff, "multidim Simulation", "linear",
"Dimensions", "Difference value")
print('Predicting on the test set using the trained RNN')
t0 = time.time()
test_predicted_time = regressor.predict(X_test)
#predicted_time = sc.inverse_transform(predicted_time)
print('Predicting on the test set complete. time_to_predict={:.2f} seconds'.
format(time.time() - t0))
# In[29]:
# Save predictions on test set
test_prediction = pd.DataFrame(test_predicted_time)
test_pred_filename = results_dir + '/' + 'test_prediction.csv'
test_prediction.to_csv(test_pred_filename)
print('Predictions on test set saved to {}'.format(test_pred_filename))
# In[31]:
# Visualize predictions on test set
test_res_plot_filename = results_dir + '/' + 'test_true_vs_pred' + '.png'
utility.plot_results(true_test_time, test_prediction,
'True vs predicted time_to_earthquake on test set',
test_res_plot_filename)
# In[32]:
# Compute error metrics on test set
test_mse = mean_squared_error(true_test_time, test_predicted_time)
test_rmse = test_mse**0.5
test_mae = mean_absolute_error(true_test_time, test_predicted_time)
print(
'Error metrics on test set. test_mse: {:.4f}, test_rmse: {:.4f}, test_mae: {:.4f}'
.format(test_mse, test_rmse, test_mae))
train_prediction.to_csv(train_pred_filename)
train_prediction_orig = pd.DataFrame(train_predicted_time_orig)
train_pred_orig_filename = params.results_dir + '/' + 'train_prediction_orig.csv'
train_prediction_orig.to_csv(train_pred_orig_filename)
print('Predictions on train set saved to {}, and {}'.format(
train_pred_filename, train_pred_orig_filename))
# In[49]:
# Visualize predictions on training set
train_res_orig_plot_filename = params.results_dir + '/' + 'train_true_vs_pred_orig' + '.png'
utility.plot_results(
y_train, train_prediction_orig,
'True (orig) vs predicted time_to_earthquake on train set',
train_res_orig_plot_filename)
train_res_plot_filename = params.results_dir + '/' + 'train_true_vs_pred' + '.png'
utility.plot_results(y_train, train_prediction,
'True vs predicted time_to_earthquake on train set',
train_res_plot_filename)
# In[50]:
# Compute error metrics on training set
N = params.ma_window
train_mse, train_rmse, train_mae, train_r2 = utility.metrics(
y_train, train_predicted_time_orig[N - 1:])
def main():
df_train = pd.read_csv('data/train.csv')
df_test = pd.read_csv('data/test.csv')
oof_logit, preds_logit = train_logit(df_train, df_test, pca=True)
ut.plot_results(oof_logit, preds_logit, df_train, 'logit_'+str(c))