Forecasting & Explainable A.I.

Background:
The usual approach of passengers, cargo and flights forecasting are not working during the social unrest and COVID age.

Solution:
With various data science techniques to aggregate large amount of data, project team was thinking out of the box to include various external data in the regular forecasting process, such as news, social media, weather, and more. With the new modelling approach, the forecasting routine should be more straight forward, while management team are notified of anomaly. Forecasting results were required to be highly explainable with improved accuracy, and support human adjustments.

Achievements:
The forecasting routine becomes a collaboration between an improved algorithm and human experts - while accuracy is improved by incorporating new data sources and features, users was alerted when anomaly happens, meaning that human intervention can be duly involved to interpret and adjust forecasting results. Meanwhile, forecasting result were fed back to the algorithm for continuous algorithm tuning and accuracy improvement.