USAID: Intelligent Forecasting Competition

United States Agency for International Development

Authors

Jonathan Hans Soeseno*, Davide Burba*, and Trista Chen (*denotes equal contribution)

Published

Nov-20

Abstract

One of our recent successes was winning first-place in an International Competition held by the US Agency for International Development (USAID). The competition focused on the global health topic, where we need to build an AI algorithm to provide an accurate prediction for contraceptive consumption in Africa, specifically Cote d’Ivoire. We used our previous experience in Inventec’s logistic forecasting to tackle the competition problem. Our winning solution consists of using an ensemble of LSTM and LightGBM, as well as informative engineered features. By employing a solid validation strategy, we tuned our solution to achieve the most accurate predictions out of approximately 80 submitted models.

Keywords

  • Global-health
  • Time-series forecasting
  • Cote d'Ivoire
  • USAID
  • Contraceptive Usage