<p>The motor insurance industry faces a persistent “vicious cycle” due to the information asymmetry, where cross-subsidization and moral hazard hinder the adoption of accurate, behavior-based pricing. Traditional usage-based insurance (UBI) models, which rely on simple event counts, often fail to capture the complexity of driving risks, limiting insurers’ ability to segment risks accurately, set fair premiums, and design incentives that promote safer driving and retention. This study proposes an innovative information system framework to break this cycle. A Driving Pattern-N (DPN) model is proposed with a novel feature set combining association-rule and sequential-pattern mining to identify causal links between driving behaviors and near-crash events. Compared to conventional UBI models, DPN captures behavioral sequences and transition probabilities, enhancing both predictive accuracy and interpretability. Implemented via a random forest ensemble, the system is benchmarked against baseline behavior-centric and pattern-based models across multiple experimental settings, demonstrating robust classification performance and actionable insights. From an information systems perspective, this research advances telematics-driven insurance by embedding transparency, explainability, and causality into the model pipeline, enabling fairer premium structures, incentivizing safer driving, and supporting a sustainable insurance ecosystem. The framework bridges UBI, advanced data mining, and decision-support systems, offering a replicable methodology for real-time risk management.</p>