Predictive Analytics for Navigation Data Using Sequence-Based Clustering and Absorbing Markov Chains

We propose a novel framework integrating absorbing Markov chains (AMCs) and sequence-based clustering (SBC) to predict and optimize absorbing behaviors in human web navigation. Unlike standard deep learning models that function as “black boxes” for next-state prediction, our AMC-SBC approach leverages interpretable matrix algebra to predict the expected remaining steps and the final absorption state. To address the heterogeneity of user behavior, we adopt SBC to cluster navigation patterns and estimate cluster-specific fundamental matrices. We demonstrate that this framework extends beyond predictive accuracy into “prescriptive analytics”. By analyzing the cluster-specific absorption probability matrices, we show how to diagnose high-impact risk states and simulate structural interventions to improve user outcomes. We validate the method using two real-world data sets, demonstrating that AMC-SBC not only competes with recurrent neural networks in classification metrics ([Formula: see text]-score, AUC) but uniquely enables granular, interpretable interventions that yield significantly higher lift than nonsegmented strategies. History: Maytal Saar-Tsechansky served