| Financial fraud has occurred around the world one after another,and the methods of fraud continue to evolve.Although the existence of China’s capital market is not long,there have been a lot of financial fraud events with a wide range of influence and harm.Although the event of financial fraud is an internal behavior of the company,the scope of influence of this behavior is by no means limited to the company.From the point of view of the whole capital market,financial fraud has seriously disturbed the efficiency and order of the capital market,and the false financial information has moved the market investors.Once the facts are revealed,the confidence of the market investors will be greatly reduced.Hesitant in the face of the next investment opportunity,thus affecting the capital flow of the whole capital market,restraining development,and doing great harm to the development of social economy.In the aspect of financial fraud identification,the early methods have great limitations and can not solve the problem of financial fraud identification perfectly.With the further development of science and technology,some researchers began to consider the application of data mining technology to the identification of financial fraud,this kind of model perfectly solved the above problems,and gradually become a more important technical means at this stage.This paper analyzes the existing research on financial fraud in data mining,and on this basis,puts forward a set of integrated model to identify financial fraud by using data mining technology.First of all,according to previous studies and self-ideas,this paper selects governance structure,external evaluation,solvency,profitability,development ability,management ability,cash flow analysis and per share index from both financial and non-financial aspects.the primary index system of the financial fraud identification model is constructed;secondly,the primary index system is screened by stepwise regression analysis,and finally six dominant variables are selected.Thirdly,three financial fraud recognition models are constructed by using python language,which are CART decision tree recognition model,SVM support vector machine recognition model and fully connected neural network recognition model,and then the Stacking integrated model is constructed on the basis of the above three fraud recognition models.Finally,An and B listed companies are selected to verify the recognition effect of the Stacking integrated identification model.Through the comparative analysis of the abnormal events,fraud causes of A company and the specific situation of B company,it is proved that the recognition effect of the model is good and can be widely used.By constructing models and analyzing actual cases,this paper verifies that Stacking integration algorithm in data mining technology has high application value in the field of financial fraud identification,which can not only assist market regulatory departments to judge key supervision and management objects,but also help certified public accountants to determine the possibility of fraud from the overall perspective of financial reporting and assist their work,and finally achieve the function of effectively assisting investors to make correct decisions.The in-depth study on the identification of financial fraud of listed companies will also help to further improve the regulatory level and supervision efficiency of the relevant institutions. |
