def ha_prediction_taxiBJ():
DATAPATH = '../data'
nb_flow = 2 # i.e. inflow and outflow
T = 48 # number timestamps per day
len_test = T * 4 * 7 # number of timestamps to predict (four weeks)
# load data
data_all = []
timestamps_all = list()
for year in range(13, 17):
fname = os.path.join(DATAPATH, 'TaxiBJ',
'BJ{}_M32x32_T30_InOut.h5'.format(year))
print("file name: ", fname)
data, timestamps = load_stdata(fname)
# print(timestamps)
# remove a certain day which does not have 48 timestamps
data, timestamps = remove_incomplete_days(data, timestamps, T)
data = data[:, :nb_flow]
data[data < 0] = 0.
data_all.append(data)
timestamps_all.append(timestamps)
timestamps_all = [timestamp for l in timestamps_all for timestamp in l]
data_all = np.vstack(copy(data_all))
print('data shape: ' + str(data_all.shape))
# make predictions
predicted_data = ha_prediction(data_all, timestamps_all, T, len_test)
# evaluate
print('Evaluating on TaxiBJ')
real_data = data_all[-len_test:]
score = evaluate(real_data, predicted_data)
# save to csv
save_to_csv('HA', 'TaxiBJ', score)
def ha_prediction_bikeNYC():
DATAPATH = '../data'
nb_flow = 2 # i.e. inflow and outflow
T = 24 # number timestamps per day
len_test = T * 10 # number of timestamps to predict (ten days)
# load data
fname = os.path.join(DATAPATH, 'BikeNYC', 'NYC14_M16x8_T60_NewEnd.h5')
print("file name: ", fname)
data, timestamps = load_stdata(fname)
# print(timestamps)
# remove a certain day which does not have 48 timestamps
data, timestamps = remove_incomplete_days(data, timestamps, T)
data = data[:, :nb_flow]
data[data < 0] = 0.
print('data shape: ' + str(data.shape))
# make predictions
predicted_data = ha_prediction(data, timestamps, T, len_test)
# evaluate
print('Evaluating on BikeNYC')
real_data = data[-len_test:]
score = evaluate(real_data, predicted_data)
# plot real vs prediction data of a region
# plot_region_data(real_data, predicted_data, (13,3), 0)
# save to csv
save_to_csv('HA', 'BikeNYC', score)
def ha_prediction_RomaNord16x8():
datapath = '../data'
nb_flow = 2 # i.e. inflow and outflow
T = 24 # number timestamps per day
len_test = T * 7 # number of timestamps to predict
# load data
fname = os.path.join(datapath, 'Roma', 'AllMap',
'Roma_16x8_1_ora_resize_north.h5')
print("file name: ", fname)
data, timestamps = load_stdata(fname)
# timestamps = np.array([adjust_timeslots(t) for t in timestamps])
# print(timestamps)
# remove a certain day which does not have 48 timestamps
data, timestamps = remove_incomplete_days(data, timestamps, T)
data = data[:, :nb_flow]
data[data < 0] = 0.
print('data shape: ' + str(data.shape))
# make predictions
predicted_data = ha_prediction(data, timestamps, T, len_test)
# evaluate
print('Evaluating on RomaNord16x8')
real_data = data[-len_test:]
score = evaluate(real_data, predicted_data)
# plot real vs prediction data of a region
# plot_region_data(real_data, predicted_data, (13,3), 0)
# save to csv
save_to_csv('HA', 'RomaNord16x8', score)