基于弱关联频繁模式的超限行为挖掘优化
An optimization for overload behavior mining based on weakly correlated frequent patterns

摘要 目前采用博弈分析和流量预测等模型，对未来时间段道路网货运车辆超限行为进行提前识别，取得一定的检测效果，但对具有时空动态性和迁移性的超限车辆分布挖掘仍具有局限性.根据道路网超限车辆数据特点，提出一种基于弱关联频繁模式的超限行为的挖掘优化算法，该算法采用空间弱关联频繁模式构建的超限频繁模式树，建立时间弱关联的状态转移模型，得到频繁模式的预测值.在FP-growth频繁模式挖掘算法的基础上，首次建立了超限模式挖掘与货运车辆行为数据的时空弱关联，使超限行为预测算法误差率降至6%以下，有效提高了超限行为的检测效率.
Abstract：At present, adopting game theory analysis and traffic prediction models to identify in advance overload vehicle in future has achieved certain detection effectiveness. However, it has the limitations on the spatiotemporal dynamics and migration of overload vehicle distribution mining. According to the characteristics of the overload vehicle data, this paper proposes an overload behavior mining optimization algorithm based on weakly correlated frequent patterns. In this algorithm, the spatial weakly correlated frequent pattern mining method is adopted to build the overload frequency pattern tree and the state transformation model of time weakly correlated. The predication value of frequency pattern is obtained. On the basis of FP-growth frequent pattern mining algorithm, this paper achieves the weak correlated between vehicle behavior data and overload behavior pattern mining. The error rate of overload behavior prediction algorithm is dropped to less than 6% and the detection efficiency is improved effectively.