look_back)
# Train the model
regressor.fit(X_train, y_train.ravel())
# Predict values
predicted_y_value = regressor.predict(X_test)
if label_target == label:
y_test_predictions.append(predicted_y_value)
X_test_new_row = np.append(X_test_new_row, predicted_y_value)
X_test_new_row = X_test_new_row.reshape(1,
no_of_features * look_back)
X_test = np.concatenate((X_test, X_test_new_row), axis=0)
X_test = np.delete(X_test, 0, axis=0)
# Calculate the RMSE between ground truth and predictions and
# add to the CSV file
rmse_val = rmse(y_test, np.array(y_test_predictions))
csv_file.add_row([
company_name, regressor_name,
str(len(y_test)), label,
str(rmse_val)
])
# Plot the graph comparing ground truth with predictions
plot_graph(
y_test, y_test_predictions, [i for i in range(len(y_test))],
directory + company_name + "_" + regressor_name + "_" + label +
".png")
def error(model, ratings_test):
return rmse([r - model.predict(uid, iid)
for uid, iid, r in ratings_test])
for m in range(nens):
filename = workdir+'/'+casename+'/'+'{:04d}'.format(m+1)+'/'+name+'.bin'
psik = util.read_field(filename, nkx, nky, nz)
# varens[m, :, :, :] = util.spec2grid(util.spec_bandpass(convertor(psik), krange, s))
varens[m, :, :, :] = util.spec2grid(convertor(psik))
varens = np.roll(np.roll(varens, -40, axis=1), 60, axis=2) #shift domain position for better plotting
##some error statistics
# for m in range(nens):
# varmean = np.mean(varens[m:m+1, :, :, lv], axis=0)
# rmse = util.rmse(varmean, var[:, :, lv])
# pcorr = util.pattern_correlation(varmean, var[:, :, lv])
# print('error = {:7.2f}'.format(rmse))
# print('pattern correlation = {:7.2f}'.format(pcorr))
varmean = np.mean(varens[:, :, :, lv], axis=0)
rmse = util.rmse(varmean, var[:, :, lv])
pcorr = util.pattern_correlation(varmean, var[:, :, lv])
print('error = {:7.2f}'.format(rmse))
print('pattern correlation = {:7.2f}'.format(pcorr))
sprd = util.sprd(varens[:, :, :, lv])
print('ensemble spread = {:7.2f}'.format(sprd))
cmap = [plt.cm.jet(m) for m in np.linspace(0, 1, nens)]
for m in range(nens):
out = util.smooth(varens[m, :, :, lv], smth)
ax[1].contour(ii, jj, out, clevel_highlight, colors=[cmap[m][0:3]], linestyles='solid', linewidths=1)
ax[1].contour(ii, jj, out1, clevel_highlight, colors='black', linestyles='solid', linewidths=2)
cax = colorbar_ax(ax[1])
cax.set_visible(False)
set_axis(ax[1], varname+' ensemble')
def evaluate_burgers(X_eval, U_eval, model: Scalar_PDE):
U_hat = model.predict(X_eval)
return util.rmse(U_eval, U_hat)
import xgboost as xgb
import csv
import sys
# preprocessing
print("start preprocessing ...")
dfTrain, dfRawTrain, dfAdr_train, dfIsCanceled_train, dfAdrReal_train, dfValid, dfRawValid, dfAdr_valid, dfIsCanceled_valid, dfAdrReal_valid, dfTest, dfRawTest = pp.preprocessing('../data/train.csv', '../data/test.csv')
print("preprocessing done ...")
print("start training ..")
# adr, is_canceled combined
modelList_adr_real, dtrain_adr_real, dvalid_adr_real = train.train_adr(dfTrain, dfAdrReal_train, dfValid, dfAdrReal_valid, 5, 10000)
print("training done ...")
# compute error
Ein_adr_real = util.rmse(dtrain_adr_real.get_label(), train.predict_adr_ensemble(dtrain_adr_real, modelList_adr_real))
Eval_adr_real = util.rmse(dvalid_adr_real.get_label(), train.predict_adr_ensemble(dvalid_adr_real, modelList_adr_real))
print("Ein_adr_real: ", Ein_adr_real)
print("Eval_adr_real: ", Eval_adr_real)
print("computing revenue ...")
# adr, is_canceled combined
revenue_train = train.predict_revenue_ensemble_adr_isCanceled_combined(dfTrain, dfRawTrain, modelList_adr_real, 0)
revenue_valid = train.predict_revenue_ensemble_adr_isCanceled_combined(dfValid, dfRawValid, modelList_adr_real, 1)
revenue_test = train.predict_revenue_ensemble_adr_isCanceled_combined(dfTest, dfRawTest, modelList_adr_real, 2)
# print("train_predict: ", revenue_train)
print("computing revenue done ...")
dfTrain_label = pd.read_csv('../data/train_label.csv')
train_label = dfTrain_label['label'].tolist()
# make sure we have a file
if exp_grey_img is None or exp_bw_img is None or gt_grey_img is None or gt_bw_img is None:
continue
# reshape if necessary
if exp_grey_img.shape != gt_grey_img.shape:
exp_grey_img = cv2.resize(exp_grey_img,
(gt_grey_img.shape[1], gt_grey_img.shape[0]))
if exp_bw_img.shape != gt_bw_img.shape:
exp_bw_img = cv2.resize(exp_bw_img,
(gt_bw_img.shape[1], gt_bw_img.shape[0]))
# rmse on greyscale masks
r = rmse(exp_grey_img, gt_grey_img)
cumulative_rmse += r
'''
#convert into bw
# thresholding less than
exp_l_channel[exp_l_channel < 50] = 0
# everything else to white
exp_l_channel[exp_l_channel > 0] = 255
'''
# read in and convert
exp_mask = cv2.cvtColor(exp_bw_img, cv2.COLOR_BGR2LAB)
exp_l_channel, _, _ = cv2.split(exp_mask)
gt_mask = cv2.cvtColor(gt_bw_img, cv2.COLOR_BGR2LAB)
# Predict the average for this user.
query['rating'] = ((float(users[cur_user]['total']) / users[cur_user]['count']) + prediction) / 2
else:
#print "Book not rated... f**k"
if users[cur_user]['count'] == 0:
# Perhaps we did not having any ratings in the training set.
# In this case, make a global mean prediction.
query['rating'] = mean_rating
else:
# Predict the average for this user.
query['rating'] = float(users[cur_user]['total']) / users[cur_user]['count']
book_except_count += 1
else:
#print "User not rated... f**k"
query['rating'] = mean_rating
user_except_count += 1
count += 1
print "Book Excepted " + str(book_except_count)
print "User Excepted " + str(user_except_count)
# Write the prediction file.
#util.write_predictions(test_queries, pred_filename)
print util.rmse(test_queries)
