<p>In the field of e-commerce, traditional customer segmentation models often rely on static transaction data and fail to capture the underlying temporal mechanisms of user behavior in massive and sparse datasets. To address this gap, this study proposes a unified framework that integrates value segmentation with human behavioral dynamics. First, we construct the RAIF model as an enhancement of the traditional RFM index to effectively mine user value from sparse shopping data by incorporating specific behavioral weights. Second, we utilize statistical physics to validate the behavioral distinctiveness of these RAIF segments. The results demonstrate that the proposed model improves segmentation accuracy in sparse scenarios and reveals that different value groups adhere to distinct physical laws. Specifically, we observe a unique power law characteristic comprising three stages in the distributions of intervals between events. This confirms that users of high value possess fundamentally different behavioral patterns compared to casual users. By integrating human dynamics into value mining, this study provides a novel theoretical perspective and actionable insights for dynamic personalized services.</p>