邵延富 谢大为
（广州市景心科技股份有限公司，广东 广州 510000）
摘要：针对传统的人工方式判断车主是否有意向购买车险存在效率低、缺乏预测性等问题，提出一种基于支持向量机的车险购买意向识别方法。首先，通过标准化处理与主成分分析降维，将30维的通话数据映射至10维空间，并采用欠抽样策略解决数据样本不平衡的问题；然后，利用SVM模型区分有、无意向车主。实验结果表明，SNM模型的识别召回率和误检率分别为97.9%、4.3%。该方法可为车险公司个性化服务提供技术支持。
关键词：车险购买意向识别；主成分分析；支持向量机；通话数据
中图分类号：TP311.5 文献标志码：A 文章编号：1674-2605(2024)06-0013-06
DOI：10.3969/j.issn.1674-2605.2024.06.013 开放获取
A Method of Car Insurance Purchase Intention Recognition Based on Support Vector Machine
SHAO Yanfu XIE Dawei
(Guangzhou Joysim Technology Co., Ltd., Guangzhou 510000, China)
Abstract: Aiming at the problems of low efficiency and lack of predictability in traditional manual methods of judging whether car owners intend to purchase car insurance, a car insurance purchase intention recognition method based on support vector machine is proposed. Firstly, through standardization and principal component analysis dimensionality reduction, the 30 dimensional call data is mapped to a 10 dimensional space, and under sampling strategy is adopted to solve the problem of imbalanced data samples; Then, use SVM model to distinguish between interested and uninterested car owners. The experimental results show that the recognition recall rate and false detection rate of the SNM model are 97.9% and 4.3%. This method can provide technical support for personalized services of car insurance companies.
Keywords: car insurance purchase intention recognition; principal component analysis; support vector machine; calling data
