C4W1 assignment - question about last cell

Course Q&A TensorFlow Developer Professional Sequences, Time Series and Prediction
1

Hi,
In the last cell of C4W1 assignment, I got the same problem that my mse & mae are not the same as the expected output.

Failed test case: diff_moving_avg_plus_smooth_past has incorrect first 5 values.
Expected:
[59.77063  58.743534 57.702236 56.543404 55.74462 ],
but got:
[40.574306 44.581184 48.20017  51.96808  55.881516].

Failed test case: diff_moving_avg_plus_smooth_past has incorrect last 5 values.
Expected:
[29.718428 33.570488 37.043247 41.25437  45.228832],
but got:
[25.880426 26.018795 26.203758 25.86561  25.94249 ].

I calculated smooth_past_series as moving_average_forecast(SERIES[SPLIT_TIME-365-10:-365], 10).
After looking at this question, it seems that I need to use SERIES[SPLIT_TIME-365-5:-365+5] instead. I understand that window size is 10 hences the parts-365-5 then -365+5, however, is it more correct to use -365-10 and -365 because the past series starts from SPLIT_TIME-365 and we need to subtract 10 for the first window?

Thank you

2

We’re perofrming centered moving average in the assignment. So, half the window size on either side of the center point is the way to go. Here’s a link you’ll find useful.