1. Business Problem
In 2017, a new regulation about invoice donation was issued. All the invoices issued will be
transformed into e-invoice instead of paper invoice. However, this regulation will cause great
donation drops, and this is a problem for central and local government in Taiwan.
This project provide forecast of the next 2 month e-invoice donation amount to central
government. With this forecast, central government could know which city or county would
confront a donation amount drop in the future and ask local government to take actions.
2. Data
We get the e-invoice donation monthly amount during 2013/1~2016/9(45 months) in top tier
(Taipei, New Taipei), middle tier (Yunlin County, Hsinchu City), bottom tier (Hsinchu
County, Chiayi County) of 6 city or county from Ministry of Finance, e-invoice platform. We
also collect the monthly promotion event in each city as the external information for some
forecasting methods .
3. Forecasting Solution
We use simple naive forecast as a benchmark. Then comparing it with different method by
validation MAPE. After we check time series & residual to avoid over-fitting, we finally
decide our forecasting method that top and bottom tiers use moving average, and middle tier
use naive forecast.
4. Recommendations:
In this report, we tend to create a technical solution for the government to control the donation
of each cities in Taiwan. We recommend the central/local government to make policy
decision according with forecasting methods in this report by different tiers of cities in
Taiwan. By looking at the threshold, they can easily read whether the donation amount is
enough and understand the e-invoice donation by not just analyzing the past data but also the
future forecast to have some reference in future policy formulation.