Personalize E-Commerce Product Recommendations Based on User Behavior Using Reinforced Learning Algorithms

The development of a personalized and adaptive e-commerce product recommendation system will be developed using the Reinforcement Learning algorithm in this study. Initial data is extremely promising in its ability to raise sales conversion: 30% of the products added to the cart are never purchased. Additionally, there is a strong correlation of 0.8 between viewed versus purchased products. Data collection was from 447 Indonesian respondents over a period of June to July 2024. It was collected using an online questionnaire that measures recommendation quality, satisfaction, and ease of use with purposive sampling. Partial Least Squares Structural Equation Modeling was done on the data analysis. From that, it has been found that system quality is positively related to the accuracy, novelty, and diversity of the recommendation. The results further show how this would lead to an improved user experience, satisfaction, and sales conversion with the reinforcement learning-based system. These findings give insight into developing efficient adaptive recommendation systems on e-commerce platforms and open opportunities for further research.本研究将采用强化学习（Reinforcement Learning）算法开发一款个性化自适应电商产品推荐系统。初步数据显示其在提升销售转化率方面表现出极强的潜力：有30%加入购物车的商品最终未被购买。此外，商品浏览行为与购买行为间存在0.8的强相关性。本次数据采集于2024年6月至7月，共收集447名印尼受访者的有效样本，数据通过在线问卷完成采集，问卷用于衡量推荐质量、用户满意度与易用性，抽样方法为目的性抽样（purposive sampling）。本研究采用偏最小二乘结构方程模型（Partial Least Squares Structural Equation Modeling）开展数据分析，结果显示系统质量与推荐结果的准确性、新颖性及多样性呈显著正相关关系。进一步的分析表明，该基于强化学习的推荐系统可有效优化用户体验、提升用户满意度并改善销售转化率。本研究结论可为电商平台高效自适应推荐系统的开发提供理论参考，同时为后续相关研究开辟了新的探索方向。