基于随机森林的保险客户利润贡献度研究

Measurement of Customer Profitability of Insurance Company in China Based on Random Forest

本文通过引入责任准备金,提出了新的保险客户利润贡献度公式,综合考虑了历史购买行为和未来可预见的现金流,更有效地度量客户的真实贡献。此外,本文首次把非参数随机森林回归法应用到保险客户利润贡献度预测中,并和其他模型进行比较,发现非参数随机森林方法往往要优于传统的类神经网络、CART和SVC等模型。实证研究发现:利用客户的年龄、性别、地区、职业、婚姻状况等变量能较准确地预测客户利润贡献度,避免了复杂的精算过程,同时还能评估潜在客户的利润贡献度。

This paper proposed a new customer profitability formula for the insurance industry by adding liability reserve. Considering the historical purchasing behavior and future cash flow for the foreseeable, it can measure effectively the customer true contribution. In addition. This paper firstly apply nonparametric random forecast regression method to the forecast insurance customer profitability, comparing with other models, finding that random forest method is often superior to neural network, CART, and SVC. Empirical study finds that Using customer＇s age, gender, region, occupation, marital status variables can accurately predict the insurance customer profitability, avoiding the complex actuarial process. Meanwhile, the model can also assess the potential customer profitability.