Optimisation and cross-validation of an e-hailing hybrid pricing algorithm using a supervised learning classification and regression tree model: a heuristic approach

Abstract

Ride-hailing application price decisions are frequently considered biased and therefore are receiving more attention. Electronic (e)-hailing providers use the machine learning reinforcement model (RL) to build their dynamic pricing (DP) strategy to charge consumers. Nevertheless, the associated pricing issues in e-hailing potentially jeopardise this flourishing industry with a more significant long-term effect if left unresolved. Upon increased demand, the DP strategies assist the e-hailing systems with price surging, the unreasonable surging of price is unexpected by e-hailing users and deters them from using e-hailing services. A drawback of the existing RL DP algorithm is that does not consider surrounding factors before surging the price. Hence, this study aimed to address the underlying pricing issues through a hybrid pricing model using classification and regression tree (CART) supervised learning. A hybrid pricing algorithm was developed to demonstrate the enhanced model has edge over the existing model. The current e-hailing RL based algorithm was compared against the SL’s CART enhanced pricing algorithm with cross-validation using centrality analysis results. The test results shows that the hybrid pricing algorithm could address DP pricing issues by optimising e-hailing prices to provide its consumers with impartial pricing and remuneration. The proposed model can be a good theoretical reference for further studies which can also be applied to other industries such as Airlines, Tourism where DP is in used.

Keywords

• Machine Learning
• Supervised Learning
• Reinforcement Learning
• CART Algorithm
• Dynamic Pricing Algorithm
• Optimised Pricing Algorithm