Optimal Clustering of Products for Regression-Type and Classification-Type Predictive Modeling for Assortment Planning

In collaboration with a national retailer, this study assessed the impact on sales prediction accuracy by including sparse demand product clusters to identify best prediction scenarios, whether regression-problem or... [ view full abstract ]

In collaboration with a national retailer, this study assessed the impact on sales prediction accuracy by including sparse demand product clusters to identify best prediction scenarios, whether regression-problem or classification-problem would lead to the best demand decision-support. Sparse demand products, where the response variable is plagued with zeros, can help obtain the probability of a product sale within a specified planning horizon so the retailer can model it in a regression setting. In our study we clustered products using several features and found a consistent story behind the clusters. Next, we aggregated the clusters into a single input feature, which led to improved prediction accuracy of the predictive models. When forecasting sales, we investigated several regression- and classification-type models and reported the best performing models. Lastly, we identified scenarios whether classification or regression approach would be more strategic for the partner and how the forecasts can help in assortment decision.