Insurance Fraud Identification Based on XGboost

• Jie Huang

In China, motor vehicle insurance plays a very important role in property insurance, and its operating status is closely related to the operating stability of property insurance companies. The growing fraud problem has many adverse effects on other honest insureds and insurance companies. The fraud identification model can identify high-risk insured persons, so that the insurer can identify suspicious cases more quickly and accurately through this model in practical work. However, due to some special characteristics of the auto insurance claim data set, such as unbalanced data distribution and various types, it has not yet been able to build a fraud identification model that adapts to a wide range. In view of the above problems, this paper cleans the data according to the characteristics of imbalance and high feature dimension of auto insurance data, and selects the features with strong prediction ability by filtering feature selection. Then, on the basis of comparing various models, this paper chooses the best-performing XGboost model to construct the insurance fraud identification model and evaluate the model. The analysis found that the indicators of XGboost are high, indicating that the performance of the model is good and can be used to accurately identify insurance fraud. This provides a reference for the insurance company platform to provide users with insurance products.

Select Volume / Issue:

• Volume 10, Issue 4, July 2023

Year:

2023

Type of Publication:

Article

Keywords:

Insurance Fraud Identification; Filtered Feature Selection; Model Comparison; XGboost

Journal:

IJASM

Volume:

10

Number:

4

Pages:

42-51

Month:

July

ISSN:

2394-2894