Analyzing consumer choices through collaborative filtering and its applications in marketing

Abstract

As the online and offline markets mature and services become more sophisticated, consumer behaviors in product purchasing have become increasingly complex. Understanding and predicting consumer behaviors are critical in achieving enhanced sales revenue through refined targeting in both online and offline markets. As one of the leading unsupervised learning methodologies, collaborative filtering can provide marketing implications through its recognition of consumers' buying patterns. The first essay suggests that collaborative filtering models predicting digital music downloads can be improved by incorporating users' tendency to seek variety, which has been under-explored in previous studies of entertainment marketing. This result may be because consumers who are more likely to search for variety tend to be influenced by product recommendations or social media promotions. The second essay presents a novel unsupervised learning technique that captures offline shopping patterns and provides a comprehensive understanding of customers’ cross-purchasing behaviors by introducing a collaborative filtering model integrated with Restricted Boltzmann Machines (RBM). While recent generative AI models have gained attention, unsupervised learning has been criticized for its lack of interpretability in results. This research offers multiple methodological approaches from a marketing perspective to address these limitations.