search

jlivsey / UB-sping24-time-series

2 stars 1 forks source link

Simantini Ghosh #13

Open Simantini1709 opened 7 months ago

Simantini1709 commented 7 months ago

time-series-analysis.pdf

vgunturi commented 7 months ago

@Simantini1709 Can you shed some light on why you did resampling in the data before forecasting?

Also, when professor implemented the prophet model on ICNSA data, the forecasted value was far away from the original one. Did you get a chance to compare the forecasted value with the original value?

Simantini1709 commented 7 months ago

To check the annual average claims and to plot it as it makes more sense to check the spike in claims in the covid year.

For the second point, it was due to winsorization.

vgunturi commented 7 months ago

Winsorization was used to treat the outliers and forecasted value was close to the actual? What's the difference? Can you post the value?

Simantini1709 commented 7 months ago

https://github.com/Simantini1709/Fred_ICNSA_FbProphet/blob/main/Time%20Series%20Analysis.ipynb

Simantini1709 commented 7 months ago

Released forecast was 261,029 .Error is my forecast is 302811.86-261029 = 41,782.86