Deliver a scalable end-to-end pipeline that makes 1 year future forecasts of 1000+ financial KPIs.
The time series analysis consists of:
- Time Series Decomposition
- Anomaly Detection
- Statistical tests to check normality, stationarity, ACF/PACF, etc.
- Fit 30+ Time Series models (Arima, Prophet, ETS, etc.), compare their performance, selects best one
- Make future forecast for all KPIs with a MAPE (Mean Absolute Percentage Error) < 5%
- Multivariate modeling
Published:August 12, 2022