Simula@BI: Predicting service time to improve route optimization
Speaker: Tarjei Bondevik, Data Scientist, Oda
The majority of an Oda driver’s workday is not spent on driving; it is spent on everything else, such as parking, re-stacking the car, and carrying goods to the customer. We define this as service time, and because of its large share of the workday, predicting it accurately becomes a business critical issue.In this presentation, I will show how we use lightgbm, a gradient boosting model, to predict the service time per customer. This leads to better route optimization, fewer delayed deliveries, and hopefully a more predictable and less stressful workday for our drivers.