KDD 2023 - Uncertainty-Aware Probabilistic Travel Time Prediction for On-Demand Ride-Hailing at DiDi скачать в хорошем качестве

KDD 2023 - Uncertainty-Aware Probabilistic Travel Time Prediction for On-Demand Ride-Hailing at DiDi

Wenzhao Jiang, Hong Kong University of Science and Technology (Guangzhou) Join us in this exciting video as we unveil DiDi's groundbreaking solution to the travel time underestimation issue in the ride-hailing industry. As one of the world?s largest ride-hailing platforms, DiDi answers billions of Travel Time Estimation (TTE) queries per day. However, millions of them suffer from underestimation issue, resulting in frustration and anxiety for passengers. In this video, we proudly introduce ProbTTE, our cutting-edge probabilistic TTE framework. This approach goes beyond mere travel time prediction by empowering the backbone deep learning model to capture the underlying distribution of TTE. By further leveraging the model's distributional output, DiDi platform gains the ability to make informed decisions on order dispatching, while users can decide whether to accept the dispatched ride considering its uncertainty. Extensive offline and online evaluation on large-scale real-world data demonstrates ProbTTE's effectiveness and utility on both travel time estimation and order dispatching services. For more details, please refer to our paper at KDD 2023. Prepare to be amazed!