Title: Learning Personalized Utility Functions for Drivers in Ride-hailing Systems Using Ensemble Hypernetworks Show Chinese title

Title: 使用集成超网络为网约车系统中的驾驶员学习个性化效用函数

Abstract: In ride-hailing systems, drivers decide whether to accept or reject ride requests based on factors such as order characteristics, traffic conditions, and personal preferences. Accurately predicting these decisions is essential for improving the efficiency and reliability of these systems. Traditional models, such as the Random Utility Maximization (RUM) approach, typically predict drivers' decisions by assuming linear correlations among attributes. However, these models often fall short because they fail to account for non-linear interactions between attributes and do not cater to the unique, personalized preferences of individual drivers. In this paper, we develop a method for learning personalized utility functions using hypernetwork and ensemble learning. Hypernetworks dynamically generate weights for a linear utility function based on trip request data and driver profiles, capturing the non-linear relationships. An ensemble of hypernetworks trained on different data segments further improve model adaptability and generalization by introducing controlled randomness, thereby reducing over-fitting. We validate the performance of our ensemble hypernetworks model in terms of prediction accuracy and uncertainty estimation in a real-world dataset. The results demonstrate that our approach not only accurately predicts each driver's utility but also effectively balances the needs for explainability and uncertainty quantification. Additionally, our model serves as a powerful tool for revealing the personalized preferences of different drivers, clearly illustrating which attributes largely impact their rider acceptance decisions. Show Chinese abstract

Abstract: 在网约车系统中，司机根据订单特征、交通状况和个人偏好等因素决定是否接受或拒绝乘车请求。 准确预测这些决策对于提高系统的效率和可靠性至关重要。 传统模型，如随机效用最大化（RUM）方法，通常通过假设属性之间的线性相关性来预测司机的决策。 然而，这些模型常常不足，因为它们未能考虑属性之间的非线性交互，并且无法满足个别司机的独特个性化偏好。 在本文中，我们开发了一种使用超网络和集成学习来学习个性化效用函数的方法。 超网络根据行程请求数据和司机资料动态生成线性效用函数的权重，从而捕捉非线性关系。 在不同数据片段上训练的超网络集成进一步通过引入受控的随机性提高了模型的适应性和泛化能力，从而减少过拟合。 我们在真实世界的数据集中验证了集成超网络模型在预测准确性和不确定性估计方面的性能。 结果表明，我们的方法不仅能够准确预测每位司机的效用，还能有效平衡可解释性和不确定性量化的需求。 此外，我们的模型作为一种强大的工具，能够揭示不同司机的个性化偏好，明确说明哪些属性对他们的乘客接受决策影响较大。