Development of a Hierarchical K-Means Clustering Recommender System Informed by Market Basket Analysis
Leveraging Word2Vec and Market Basket Segmentation to Enhance Personalized Recommendation Algorithms
Introduction
Recommendation systems are vital in the e-commerce industry, as they provide tailored product suggestions to customers based on their preferences and behaviors. While traditional recommendation methods like Collaborative Recommendation and Market Basket (Sequence) Analysis have been widely used, they do come with certain limitations.
Collaborative Recommendation primarily centers on customer behavior similarity, but it can become computationally demanding when dealing with extensive product catalogs. Moreover, it often overlooks the relationships and commonalities among products, especially when valuable information like “product descriptions” is available. Furthermore, Collaborative Recommendation typically employs clustering methods like ‘K-Means’ and ‘Linkage Clustering’ using original features such as the quantities of different products purchased as clustering variables. Unfortunately, this frequently results in suboptimal performance, with skewed segment sizes and segments that only reflect statistical purchasing patterns, failing to capture the true ‘collaborative behavior’ of customers. Consequently…
