Research On Suspicious Transaction Recognition Of Insurance Company Based On Random Forest And K-means Algorithm Model

With the continuous growth of modern economy,the complexity of financial system and the continuous growth of network technology,the problem of money laundering in various countries is constantly increasing.Money laundering is a process that relies on various financial transactions to transform illegal gains into ostensibly legal ones.Based on the actual characteristics of money laundering transactions,this kind of education has remarkable planning,secrecy and specificity.Criminals hide their illegality by making elaborate,apparently legal transactions.At present,major countries have a high degree of attention to anti-money laundering,anti-money laundering supervision has also triggered the full attention of multiple subjects,the central bank online anti-money laundering supporting regulatory measures,reasonable identification of all kinds of anti-money laundering transactions.According to the central bank’s various norms,the major financial institutions comply with the law to promote the screening and monitoring of various tasks.However,the current anti-money laundering supervision of insurance enterprises mainly relies on a large number of manual operation,which has the characteristics of relatively large task scale,relatively complex process and relatively low efficiency,and is more likely to cause misjudgment and missed judgment.With the continuous growth of AI and data mining in the new era,it relies on the latest technology to identify money laundering transactions by itself,and then turns it into a highly concerned task for major enterprises.This paper takes y property insurance company as the research object.It is a property insurance company controlled by a central enterprise.After 14 years of continuous development,it mainly provides insurance services to the power grid system and upstream and downstream suppliers.In order to meet the regulatory requirements and meet the needs of digital transformation,the anti money laundering work introduces data mining technology to accurately explore suspicious transaction clues,so as to put it on the agenda.Firstly,this paper describes the background of anti money laundering regulation and the basic research status at home and abroad,and further defines the research theme of this paper;Secondly,it defines the concepts of anti money laundering,anti money laundering risk and anti money laundering supervision,which provides theoretical support for this paper;Then it analyzes the current situation and problems of anti money laundering of Y property insurance company.in order to solve the problem that the information technology of the anti money laundering process of Y property insurance company is backward,relying heavily on the traditional rule engine judgment and the recognition rate is low,an optimized model is designed;Thirdly,in order to build a data mining model,relevant identification of money laundering transactions was carried out,and the paper’s random forest and k-means were elaborated by integrating relevant literature and theoretical research.The identification model of anti money laundering transactions is constructed by using the random forest and K-means algorithm models,and the data of Y property insurance company is used for empirical analysis,and the effects of different models and methods are compared,And compared with the current traditional rule engine process.The conclusions of this paper are as follows: random forest has relatively stable characteristics,the actual value of prediction accuracy is 100%;The actual value obtained by k-means accuracy is 39%;The actual value of the logistic regression accuracy rate is 86%,and the actual value of the traditional rule engine accuracy rate is 66%.From the perspective of operation efficiency,the time consumed by traditional engine screening significantly exceeds that of the other two models.So,rely on the case study can conclude,random forests and K-means model,the growth prediction accuracy to a certain extent,about the efficiency of contrast,random forests in test set the specific operation time for 2seconds,K-means the specific length 30 seconds,10 seconds in logistic regression,traditional rules engine determine the specific time is 4000 seconds,The actual recognition efficiency will increase significantly;Based on empirical modeling,the paper points out that insurance companies can actively optimize the anti-money laundering suspicious transaction identification mechanism and use anti-money laundering opinions in the subsequent development process.Based on the in-depth analysis of this paper,new technologies such as data mining and AI will be introduced into the prediction scenario of anti-money laundering,which will provide more important reference basis for the subsequent development and optimization.