Query-Based-Learning Genetic Algorithm to Construct Mobile-Oriented Catalogs in M-Commerce

Abstract

This paper seeks to optimize customer attraction for mobile apps in m-commerce. We model this problem as a mobile-oriented catalog (MOC) segmentation problem. We use query-based learning (QBL) to develop MOCs and aim to attract most number of customers through minimal number of MOCs. This paper illustrates how to design attractive MOCs to recommend items by using QBL genetic algorithm (QBLGA). We propose preference modeling, Product2Vec, Transaction2Vec, and their variations as our oracles of QBLGA. These oracles can aggregate similar purchasing experiences to optimize combination of products for MOC construction. We divide these oracles into major oracle and minor oracles, and then, QBLGA uses these two types of oracles to produce high-attractive products. Experiments show that QBLGA outperforms the state-of-the-art algorithm to attract the greatest number of customers.