Adding auto regression n_lags- seems to remove any predictions or cause a length mismatch in datasets? very basic-

[suntochukwu]

I feel like this is extremely basic. The data set is just gold stock market prices- ds/datetime and closing price/y.

mgc_data = yf.download('MGC=F', period='1y', interval='1h')
mgc_data= pd.DataFrame(mgc_data).reset_index()
mgc_data.rename(columns={'Datetime': 'ds' ,'Close': 'y' }, inplace=True)


#--prophet
basicdf= mgc_data[['ds','y']]
basicdf.dropna(inplace=True)`

Here is the code:

m2 = NeuralProphet(drop_missing= True , growth= 'discontinuous', n_lags=50)
m2.set_plotting_backend("plotly-static")
metrics = m2.fit(basicdf, freq= 'H')
#future = m2.make_future_dataframe(basicdf, periods=50, n_historic_predictions=True)
forecast = m2.predict(future) 

Without n_lags the models works and predicts well. When i add autoregression i get an error:

dataset returned:
ds step0 trend0 season_weekly0 season_daily0 ar0 step1 trend1 season_weekly1 season_daily1 ... step98 trend98 season_weekly98 season_daily98 ar98 step99 trend99 season_weekly99 season_daily99 ar99
0 2023-06-08 20:00:00 NaN NaN NaN NaN NaN NaN NaN NaN NaN ... NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN
1 2023-06-08 21:00:00 NaN NaN NaN NaN NaN NaN NaN NaN