Carlson School of Management
With the advent of the big-data era, recommender systems are becoming increasingly important for electronic commerce, as they assist users in finding relevant content, products, and services among the ever-increasing set of available alternatives. Although recommendation techniques have proven to be effective predictive methodologies, prior work has focused largely on modeling individual, subjective, taste-driven consumer preferences in application domains of "experience" goods (e.g., movies, music, or books), and the applicability of such methodologies in domains of highly different nature is under-explored. These researchers are interested in extending recommender systems methodologies to the domain of sales forecasting by presenting a latent-factor-based modeling technique, which applies a tensor factorization approach for time-aware, simultaneous modeling of past and current sales across multiple products and multiple stores, and utilizes a seasonal time-series model to extend prediction results to future time points. The advantages of this method can be demonstrated by comparing its performance to a number of techniques from prior literature on a large dataset of sales transactions from more than 2,000 grocery stores across 47 U.S. markets. Specifically, the new approach improves upon existing methods in terms of product sales forecasting accuracy and can provide valuable decision support (e.g., for store managers or product distributors) on new product introduction.