#data_prev = bm.merge_two_data(data_prev_2, data_prev) #data = bm.merge_two_data(data_prev, data) #channge missing values 0 to NaN data.replace(0, np.nan, inplace = True) #add one hots to data data = bm.join_weekday_one_hot(data) data = bm.join_daypart_one_hot(data) #data = bm.join_minute_one_hot(data) #Prepare the sets features = len(data.columns) x_features = features * TIME_STEP reframed = bm.series_to_supervised(data, TIME_INTERVAL, TIME_DIFFERENCE, SAMPLE_FREQUENCY) train = reframed[reframed.index.month < MAY] test = reframed[reframed.index.month == MAY] #removing weekends from database #train = train[train.index.weekday < 5] #test = test[test.index.weekday < 5] x_train, y_train = train.values[:,:x_features],train.values[:,-1] x_test, y_test = test.values[:,:x_features],test.values[:,-1] #reshape the x's to 3D[sample, time_steps, features] x_train = x_train.reshape([x_train.shape[0], int(x_train.shape[1] / features),features]) x_test = x_test.reshape([x_test.shape[0], int(x_test.shape[1] / features),features])
"Sisli": "speed_data\\Sisli\\preprocessed_1745_2017.csv", "Sultanbeyli": "speed_data\\Sultanbeyli\\preprocessed_636_2017.csv", "Umraniye": "speed_data\\Umraniye\\preprocessed_619_2017.csv" } mapes = {} for file_name in FILES.keys(): time_interval = 20 data = bm.read_data(FILES[file_name]) data['Scaled'], sc = bm.scale_data(data) data.drop(['Speed'], axis='columns', inplace=True) data.replace(0, np.nan, inplace=True) reframed = bm.series_to_supervised(data, time_interval, 7 * 24 * 60, 5) reframed = reframed[reframed.index.month == 5] est = np.mean(reframed.values[:, :-1], axis=1) result = reframed.values[:, -1] mape = bm.mean_absolute_percentage_error(result, est) mapes[file_name] = mape mapes_20 = pd.DataFrame.from_dict(data=mapes, orient='index', columns=['mape_20']) mapes = {} for file_name in FILES.keys(): time_interval = 30