Abstract: With the occurrence of highway congestion and uneven spatial and temporal distribution of traffic flow in China, the supply side reform of expressway, which took expressway differential charging as the main way, was an important measure for the diversified expressway users to alleviate the current highway congestion. Based on the customer segmentation theory, the RFMS model was put forward, in which the expressway users were divided into heterogeneous customer groups and their commercial value was analyzed. Considering the structural characteristics of expressway toll data, the small clustering phenomenon of traditional k-means algorithm in the initial cluster center selection and big data clustering was improved, a hybrid algorithm combining Adaboost and k-means + + was established, and a comparative analysis of the clustering effect was carried out. The research results show that the Adaboost-k-means + + hybrid algorithm can improve the clustering effect; the business value mining method of expressway customers based on RFMS can effectively divide heterogeneous customer groups and evaluate customer value, which can provide reliable theoretical support for expressway differential charging and expressway operation management departments.

Key words: traffic engineering, highway, RFMS, customer segmentation, commercial value, differentiated charges, Adaboost

摘要： 随着我国高速公路道路拥堵和车流时空分布不均等现象的发生，面对多元化的高速公路使用者，以高速公路差异化收费为主要方式的高速公路供给侧改革是缓解目前高速公路道路拥堵等现象的重要措施。基于客户细分理论提出RFMS模型将高速公路使用者划分为异质性客户群，并分析其商业价值。考虑高速公路收费数据的结构特点，对传统k-means算法在初始类簇中心选取和大数据聚类下出现的小聚类现象进行改进，构建Adaboost与k-means++相结合的混合算法，并对聚类效果进行对比分析。研究结果表明：Adaboost-k-means++混合算法有利于提升聚类效果；基于RFMS的高速公路客户商业价值挖掘方法能有效划分异质性客户群并评估客户价值，可为高速公路差异化收费和高速公路运营管理部门提供可靠的理论支持。

关键词: 交通工程, 高速公路, RFMS, 客户细分, 商业价值, 差异化收费, Adaboost