A path fitting accuracy improvement and evaluation method based on minimum toll and multi-dimensional anomaly detection engine

Abstract Aiming at the problems of inaccurate path fitting and “same road, different tolls” caused by missing or abnormal gantry data in expressway network tolling, this paper proposes an innovative two-stage path fitting framework. The first stage (baseline fitting) rapidly generates spatiotemporally reasonable paths based on the whole-network topology. The second stage (optimized fitting) deeply integrates the business principle of “minimum toll” with a parallelized multi-dimensional anomaly detection engine, which systematically defines six types of real-world anomalies and provides their mathematical detection models. In addition, this paper is the first to propose a dual-indicator quality evaluation model, providing a quantitative basis for the reliability of fitting results. Experiments based on real transaction data show that, compared with traditional shortest-path fitting and baseline methods, the proposed framework improves path fitting accuracy by approximately 2.1% and reduces the billing deviation rate by nearly 70%. The confidence indicator exhibits a strong negative correlation with actual fitting error (Pearson coefficient = -0.87), verifying the effectiveness of the evaluation model. Finally, case scenarios demonstrate the practical application of the two-stage framework and evaluation model. This work provides a reliable theoretical and practical solution for realizing accurate and fair expressway network tolling.