0
$\begingroup$

I have several years of historical data, which shows the number of users each hour for the 72 hours prior to tax day for a tax-prep website. I am trying to predict the number of users using FaceBook Prophet for this year, and am getting weird results:

|      ds             |   yhat  |yhat_lower|yhat_upper|
| ------------------- | ------- | -------- | ---------|
| 2021-05-14 00:00:00 | 3064.06 | -1958.27 | 7676.18  |
| 2021-05-14 01:00:00 | 4.71    | -4813.93 | 5183.06  |
| 2021-05-14 02:00:00 | -1794.45| -6718.39 | 3271.49  |
| 2021-05-14 03:00:00 | -2946.66| -7978.80 | 1913.50  |
| 2021-05-14 03:00:00 | -3775.24| -8282.93 | 1325.84  |

This is how I trained the model:

# instantiate the model and fit the timeseries
prophet = Prophet()
grouped = df.groupby("activity_tax_year_num")
for g in grouped.groups:
    group = grouped.get_group(g)
    m = Prophet()
    m.fit(group)
    future = m.make_future_dataframe(periods=365)
    forecast = m.predict(future)
    print(forecast.tail())

One issue that I'm guessing is messing with my results here is the fact that tax day has fallen on different days for the past several years. Although the dates are different, each time period shows the 72 hours prior to tax day.

I'm also confused on how to use multiple timeseries to forecast for just one year?

Any guidance would be greatly appreciated!

$\endgroup$

0

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.