A lifecycle-based household recommendation system: From product recycling to purchasing

The rapid expansion of the online recycling service market, along with the increasing availability of consumer recycling data from e-commerce platforms, presents new opportunities to enhance recommendation systems (RSs) by incorporating comprehensive lifecycle information. This study introduces a novel Product-Customer Interaction Lifecycle (PCILC) model, designed to capture the distinct purchasing and recycling behaviors of household consumers. Leveraging machine learning techniques, including Bayesian Networks and Factorization Machines (FM), we propose a three-stage recommendation framework that integrates collaborative filtering (CF) with recycling data to improve predictive accuracy. First, collaborative filtering is applied to model user preferences based on historical purchase records. Second, repurchase patterns are analyzed to capture household consumption cycles. Finally, Factorization Machines incorporate recycling data to establish associations between recycled and newly purchased products, refining recommendation outcomes. Empirical evaluation on a real-world dataset demonstrates a significant correlation between recycling and purchasing behaviors, highlighting the critical role of lifecycle data in improving RS performance for household consumers. Our proposed framework increases the F1-score by 30% compared to baseline models, marking the first attempt to systematically integrate recycling data into recommendation systems for household consumption.