Vehicle insurance rapid claim settlement method and system based on tree model

Abstract

The invention relates to a vehicle insurance rapid claim settlement method and system based on a tree model. The method comprises the steps of: processing or filling abnormal values and null values invehicle insurance claim settlement data; carrying out dichotomy on the preprocessed data to form a data set; constructing a feature library according to different feature types; based on the data setand the feature library, using an XGBoost tree model to construct a vehicle insurance rapid claim settlement model; randomly extracting risk cases and normal cases according to a certain proportion,constructing a training set, and determining the optimal proportion of positive samples to negative samples again according to training results. According to the invention, the accuracy and interpretability of the model are significantly improved, and the efficiency of vehicle insurance claim settlement is greatly improved.

Description

technical field [0001] The invention relates to the technical field of auto insurance claim automatic discrimination algorithm, in particular to a method and system for fast auto insurance claim settlement based on a tree model. Background technique [0002] According to the statistics of relevant departments, small-amount cases account for more than 70% of the auto insurance claims business, and the claim settlement cycle still takes about 1 to 2 weeks, and the entire process is very time-consuming. Based on cost control and customer service considerations, insurance companies expect to quickly settle claims and close cases for small-amount cases to reduce manpower investment. However, in the field of auto insurance, there are a lot of fraudulent activities. Balancing the time requirements of the claim settlement process with the insurance company's anti-fraud risk control requirements is the core requirement of the rapid claim settlement business. The traditional auto ins...

AI Technical Summary

Problems solved by technology

Due to the different professional levels of experts, it is very likely that the ability to identify hidden risks in some cases is insufficient, and some risk factors may be missed

The more important point is that the rule engine is easy to be breached, and once it is guessed by hackers, it will fail, which greatly reduces the security and effectiveness of the rule model

Traditional models such as logistic regression and support vector machines have fewer methods for processing sparse and unbalanced data such as auto insurance fraud. The learning ability of the model itself is limited, and it is difficult to find the risk factors behind fraud cases, which cannot meet the high-precision requirements of the business.

Images