The materials in this repository are companion pieces to an article describing the predictive models for tax revenue forecasting based on official tax data from kgd.gov.kz.
Vladimir Kolesnikov — Senior Data Scientist at Sber
Abylaykhan Yergesh — Junior Data Scientist at Sber
Valeriya Rudikova — Junior Data Scientist at Sber
Python 3.8 version or above
silverkyte==0.2.0
fbprophet==0.7.1
Other installation options are described in the documentation.
The full documentation for silverkite available at https://github.com/linkedin/greykite.
The full documentation for fbprophet available at https://github.com/facebook/prophet.
| model name | MAPE | R2 | q_test(a=0.05) | q_test_pvalue |
|---|---|---|---|---|
| arima | 7,859 | 0,934 | 0,007 | 0,93 |
| fbprophet | 5,668 | 0,953 | 1,21 | 0,269 |
| silverkite | 3,958 | 0,984 | 0,095 | 0,756 |