Online Information Serves Offline Sales: Knowledge Graph-Based Attribute Preference Learning
提出一种基于知识图谱的多属性偏好学习方法（KG-APL），利用线上数据学习消费者属性偏好，帮助线下销售人员制定个性化营销方案，并通过实验验证其有效性。
The combination of online and offline shopping is becoming more common. We define the scenario where online information serves offline sales as online–offline scenario. Online and offline shopping each has its own advantages and challenges, making it necessary to integrate the strengths of both online and offline channels. Online platforms can discover consumers' attribute preferences through their online generated data. Since consumers maintain coherent attribute preferences over a period of time in both online and offline, offline salesmen can use the attribute preferences transferred from online platforms to create personalized marketing plans. In this online–offline scenario, it is crucial to learn consumer attribute preferences. We propose a knowledge graph-based multiattribute preference learning method (KG-APL), which integrates knowledge graph (KG) and multiattribute decision-making (MADM) theory. Based on MADM theory, KG-APL can learn multilevel attribute preferences in a data-driven way and provide an explanatory analysis for attribute preferences. The explanations rely on both the MADM theory and rich side information about product contained in the KG. Specifically, MADM describes the consumer's decision-making process and KG provides a hierarchical structure from products to attributes and subattributes. To verify its effectiveness and robustness, we use randomly generated data for experiments and real-life data for simulated decision making. Our article provides insight into the way of achieving integration between online and offline channels and offers theoretical and methodological support in enhance online–offline purchase services.