Predicting Offers for Mobility-on-Demand Services: A Machine Learning-Based Approach

The rapid development ofmobile internet technology has driven the widespread adoption of Mobility on-Demand services like ride-hailing and ride-pooling, which are integral to daily travel. Ensuring the quality of these services is vital for Transportation Network Companies such as Uber, Lyft, and Didi, as it directly impacts user experience and revenue. One quality aspect is providing timely and precise predictions of waiting and driving times for users, as this information affects their decision-making. However, creating these predictions is conceptually and computationally challenging in high-demand scenarios due to the dynamic nature of the services and exhaustive vehicle information processing. This paper introduces a machine learning-based model incorporating a three-dimensional encoding method based on pick-up likelihood, vehicle capacity, and detour degree to efficiently encode user requests and fleet status. A selection mechanism filters encoded information in temporal and spatial domains, reducing computational burden. The model employs an encoder-decoder architecture, with a Convolutional Neural Network encoding fleet status as an image and Fully Connected Networks decoding it for various predictions. A case study using the Manhattan taxi dataset demonstrates the model’s effectiveness, showing that spatial domain information significantly impacts predictive performance. The model achieves a True Negative Rate of approximately 90% in identifying unserviceable requests and outperforms heuristic algorithms in travel time prediction accuracy and efficiency, though it lags in waiting time prediction. Future research should explore nonlinear encoding for vehicle pick-up likelihood and test various neural network configurations to improve prediction performance and transferability.     «

The rapid development ofmobile internet technology has driven the widespread adoption of Mobility on-Demand services like ride-hailing and ride-pooling, which are integral to daily travel. Ensuring the quality of these services is vital for Transportation Network Companies such as Uber, Lyft, and Didi, as it directly impacts user experience and revenue. One quality aspect is providing timely and precise predictions of waiting and driving times for users, as this information affects their decisio...     »