1

This is the image of plot components [][1]

This is the image of forecasted value[][2]

Orange line shows yearly plot [][3]

Links to images are:[1]: https://i.stack.imgur.com/ATy2i.png [2]: https://i.stack.imgur.com/dSSzL.png [3]: https://i.stack.imgur.com/VWJZz.png

According to prophet my forecast should have been made of yearly component plus trend. However my yearly component is nothing like what plot component showed yearly seasonality as .
Can someone please explain how and why my fit or yearly plot not similar to plot_components yearly . And how exactly is my yearly plot fitted(as show) cause it seems like it is overfitting?

deluxe_daily_proph=deluxe.reset_index()
m2=Prophet()
m2.fit(deluxe_daily_proph)
future2=m2.make_future_dataframe(periods=50,freq='M',include_history=True )
forecast2=m2.predict(future2)
forecast2[['ds','yhat','yhat_lower','yhat_upper']].tail()
plt.plot(forecast2['ds'],forecast2['yhat'])
plt.plot(deluxe_daily_proph['ds'],deluxe_daily_proph['y'])enter code here
m2.plot_components(forecast2);
Shreya Sinha
  • 11
  • 5
  • So your issue is that your actual plot does not look like the sum of 'trend' and 'yearly' values? It does look like the 'yearly' plot in image components is overfit, but I suggest you test your correlation coefficient to be sure. Please edit your question to be more clear, and if possible, show your code so that we can understand the regression that you've applied. – ChaddRobertson Aug 18 '20 at 09:11
  • My question here is basically shouldn't yearly plot of plot_components and the yearly plot (3rd image ) should have a similar value as both represent seasonality in a year. How is the exactly fitting to be different every year ? @ChaddRobertson – Shreya Sinha Aug 18 '20 at 10:17

0 Answers0