CODEWITHSUNDEEP
pythonfacebook-prophet
Jeffery Wong
Jeffery Wong

Reputation: 11

Can't use prophet for forecasting ( Dataframe has less than 2 non-NaN rows)

My notebook in kaggle

Please focus on the block near the bottom of my notebook. I can't do a forecast with an error message "ValueError: Dataframe has less than 2 non-NaN rows."

What Can I do to resolve it ???

tic = time.time()

for s in proph_results['shop_id'].unique():
    for i in proph_results['item_id'].unique():
        proph_train = train.loc[(train['shop_id'] == s) & (train['item_id'] == i)].reset_index()
        proph_train.rename(columns={'date': 'ds', 'item_cnt_day': 'y'}, inplace=True)

        m = Prophet()
        m.fit(proph_train[['ds', 'y']])
        future = m.make_future_dataframe(periods=len(test_old.index.unique()), include_history=False)
        fcst = m.predict(future)

        proph_results.loc[(proph_results['shop_id'] == s) & (proph_results['item_id'] == i), 'sales'] = fcst['yhat'].values

        toc = time.time()
        if i % 10 == 0:
            print("Completed store {} item {}. Cumulative time: {:.1f}s".format(s, i, toc-tic))

Upvotes: 0

Views: 8824

Answers (1)

Ganesh Jadhav
Ganesh Jadhav

Reputation: 802

Prophet cannot be used when number of rows (which are not null) in the data you're passing is less than 2. So, you cannot do the prediction in that case.

So you get an error when you're fitting the model.

There's no other solution other than adding more (not null) data to the existing.

Upvotes: 8

Related Questions