Prediction and control of delivery times in service platforms with Artificial Intelligence

Abstract

Delivery service platforms, or delivery platforms, play an increasingly important role in the consumption of different products and services. This work uses data from a platform for the sale and distribution of cooking gas to end consumers to build a machine learning model to predict delivery times for use in service management. Different machine learning models are applied and compared, and the analysis allows the identification of the main offenders for better control of delivery times and business decision-making in the operation of the platform. The final model uses the Extremely Randomized Trees algorithm and is more accurate than the method currently employed by the company.