关联规则向量化挖掘算法及其在车险精算中的应用

Vectorization Data-mining Algorithm of Associate Rule and Its Application in Auto Insurance

本文首先回顾了关联规则的基本概念和传统的Apriori算法,然后利用关联规则的数据库是布尔型数据库的特点,在计算关联规则的支持度和置信度的时候引进向量数乘和向量内积的概念,得到关联规则向量化挖掘算法ARVDA,避免全数据库逐条记录模式匹配和属性分层,提高算法的速度．最后,本文利用提升度量关联规则的重要性,采用单独追踪和对比分析方法,衡量车险精算中风险因子的有效性．结果表明把车辆使用性质作为车险定价的分级因素是比较合理的,对于非运营车辆需要合理的费用附加．

This paper first reviews the basic concept of associate rule and Apriori algorithm. Since the values of objective dataset are binary, ＂number＂ product and inner product of vectors are introduced to develop associate rule vector data-mining algorithm （ARVDA）. It avoids matching the pattern one record by one record in the whole dataset and accelerates the computation. Finally, the method is used in the data mining of traffic accidental information to search the patterns of traffic accidents and provide risk analysis for auto insurance product pricing. Results show that the purpose of using auto should be considered as an important factor to efficiently distinguish risk levels of auto insurance.