Integrating Acceptance Prediction, Assignment, and Pricing Optimization to Maximize the Expected Revenue of Ride-Hailing Transactions

Main Article Content

Eka Kurnia Asih Pakpahan, Andi Cakravastia, Anas Ma’ruf, Bermawi Priyatna Iskandar

Abstract

The ride-hailing platform mediates transactions between driver and passenger through an assignment algorithm. However, more than 40% of assignments are rejected either by drivers or passengers, leading to inefficient system performance. Predicting which assignment would be accepted becomes necessary to avoid such inefficiency. Big data technology and the availability of ride-hailing daily transactional data can enable platforms to dive into driver’s and passenger’s behavior and determine factors influencing their acceptance. This paper aims to incorporate prediction ability into an assignment optimization model. We perform numerical experiments and conclude that the proposed model could provide behavior-custom assignment, thus ensuring a high acceptance rate by drivers and passengers. Unfortunately, We found that the global optimal solution for the model can only be found for small-sized problems. When dealing with large-sized problems, standardized optimization software can only provide local optimal or no feasible solution. We designed an assignment algorithm based on particle swarm optimization to improve the solution-finding process. The algorithm can achieve an average of 1.78% error for small-sized problems, up to 6.47% better solution compared to local optimal solutions found by standard optimization software, and it can give feasible solutions when standard optimization software fails to provide them.

Article Details

Section
Articles