A Utility-Mining-Driven Active Learning Approach for Analyzing Clickstream Sequences

In the rapidly evolving e-commerce industry, the ability to select high-quality data for model training is essential. This study introduces the High-Utility Sequential Pattern Mining using SHAP values (HUSPM-SHAP) model, a utility mining based active learning strategy, to tackle this challenge. We found that the parameter settings for positive and negative SHAP values affect the mining results of the model, introducing a key consideration into the active learning framework. Unlike traditional SHAP, which evaluates individual elements, HUSPM-SHAP utilizes SHAP values in combination with HUSPM to identify the valuable of high-utility sequential patterns for improving prediction models. In experiments to predict behaviors that actually lead to purchases, the developed HUSPM-SHAP model shows its superiority in different scenarios. The model’s ability to reduce labeling requirements while maintaining high predictive performance is highlighted. Our results show that the model is able to refine the processing of e-commerce data and lead to optimized, cost-efficient prediction modeling.