search

kiccho1101 / kaggle_m5_forecasting

My 207th🥈 solution for Kaggle M5 Forecasting Accuracy Competition
https://www.kaggle.com/c/m5-forecasting-accuracy
7 stars 1 forks source link

ARIMA #23

Closed kiccho1101 closed 4 years ago

kiccho1101 commented 4 years ago 
predictions = []
for row in tqdm(train_dataset[train_dataset.columns[-30:]].values[:3]):
    fit = sm.tsa.statespace.SARIMAX(row, seasonal_order=(0, 1, 1, 7)).fit()
    predictions.append(fit.forecast(30))
predictions = np.array(predictions).reshape((-1, 30))
error_arima = np.linalg.norm(predictions[:3] - val_dataset.values[:3])/len(predictions[0])