A Recommendation Algorithm for Auto Parts Based on Knowledge Graph and Convolutional Neural Network

• Published in: Big Data Vol. 1709; pp. 57 - 71
• Main Authors: Lin, Junli, Yin, Shiqun, Jia, Baolin, Wang, Ningchao
• Format: Book Chapter
• Language: English
• Published: Singapore Springer 2022; Springer Nature Singapore
• Series: Communications in Computer and Information Science
• Subjects: Auto-parts; Data mining; Graph Convolutional Network (GCN); Knowledge graph (KG); Recommendation
• ISBN: 9811983305; 9789811983306
• ISSN: 1865-0929; 1865-0937
• DOI: 10.1007/978-981-19-8331-3_4

Due to the large number and complex types of auto parts, the ability that algorithm can accurately match the right auto parts is a major problem for buyers to solve. To address this problem, we propose a knowledge graph convolutional network with user history and item entity augmentation for auto parts recommendation system. First, the knowledge graph of auto parts and the knowledge graph of users to find a set of node examples. Second, collect contextual data about each instance. This data is to obtain information such as the hierarchical type, role, attribute value, and inferred neighbors of each neighbor instance from the knowledge graph. Finally, process this information into an array and use this array as the input to the neural network. After one or more aggregations and merging, as a vector of construction examples, the vector contains a wide range of information. The algorithm combines the advantages of knowledge graph and graph convolution, and combines with the idea of recommendation based on item content and recommendation based on user, so that the recommendation effect is improved.