- Pre-processed client's historical data from 2019 to 2022 by Pandas, visualizing and converting data into the stationary
- Cross-validated and tuned the ARIMA model through forward-chaining to final optimal parameters, achieving the lowest MAPE of ARIMA on the test set
- Visualized the time series output by Python, proactively communicating with the client to help them to reach the optimal budget allocation, potentially saving ads spending by 20