Optimisation and Cross-Validation of Hybrid Pricing Algorithm using SL’s CART Model for E-Hailing

Abstract

Machine learning (ML) transforms the world and creates new technological avenues. The ML models such as supervised, unsupervised, and reinforcement learning (RL) offer various simple, medium, and complex solutions to the real world. Banking, transportation, and e-commerce rely on ML models. The e-hailing falls under the transport industry which relies heavily on ML’s RL model to build its dynamic pricing (DP) strategy. However, associated pricing issues due to limitations of the RL model could potentially jeopardize e-hailing grandeur and impact the revolutionary sector. The DP assists e-hailing providers with surging prices during peak hours or high demand, which many consumers believe is discriminatory pricing for riders and drivers. This paper aims to explain and highlight how to address underlying issues caused by the RL model with supervised learning’s classification and regression tree (CART) model by constructing a hybrid pricing model. The prototype will demonstrate the potential to enhance the pricing model and downgrade pricing issues. The e-hailing is an essential sector of Malaysia that will remain relevant. If issues arising from RL’s dynamic pricing are unaddressed, it can create a more significant impact. Hence, the researcher initiated a study to prove that the proposed CART model has an edge over the existing RL model of the transportation industry. The test result shows improved pricing offered by the hybrid pricing model. Further, the solution can act as good research for further related to dynamic pricing and a similar solution can be extended to tourism, airlines, or other industries that utilize the RL model to offer dynamic pricing.