Recognition On Profit Manipulation In Chinese Listed Companies Based On The Improved SVM Model

With the development and progress of China's securities market, China's listed companies are increasingly becoming an important force in the economy running, and among them, profit as the core indicators of business performance, but also has more and more important effect in the listed company and role. However, at present, because the information disclosure system of China's securities market is not perfect, means of government regulation and regulatory experience is insufficient, cause of information asymmetry between investors and institutions, and for some bad listed company provides space and opportunities to undertake profit manipulation. Also because of the profit manipulation of listed companies, cause the accounting information distortion, and influence the enthusiasm of investors entering the securities market, finally make a very adverse impact to the healthy development of the securities markets. Therefore, strengthen the management of profit manipulation has become the attention of the government administration and the broad masses of investors.In the management of profit manipulation, accurate recognition of profit manipulation The key factor, especially as the Shanghai port on April 10 th, 2014 formally approved to be trying, the relevance of Chinese securities market and the international securities market is more and more strong, then the supervision difficulty become bigger and bigger. Only timely recognition of profit manipulation of listed companies, the market can be effectively maintain and the economic can stably develop. Therefore, there are important theory value and practical significance for maintaining the investment interests of interests related groups and promoting fair and specification of China's securities market trading to build scientific and effective profit manipulation recognition model of Chinese listed companies.Based on the above analysis, the study find China's listed companies as the research object. Firstly, on the sample selection, the study collected 2421 sample companies from the a-share listed companies with eliminating financial and insurance companies and missing value samples, including 119 profit manipulation sample companies and 2302 no profit manipulation sample companies. Secondly, on the index selection, the study according to the means of profit manipulation and financial statements chose 14 external operating indicators, and using factor analysis method reducing to six main factors, at the same time according to the fraud triangle theory to analysis of the enterprise's pressure, opportunity and self rationalization factor of profit manipulation behavior and chosing 10 internal motive indicators, and using independent sample t-test analysis mean screening six indexes which existing significant difference characteristics. Thirdly, on the choice of model, the study choose the nonlinear support vector machine(SVM) intelligent technology to recognize profit manipulation, but at the same time introducing mixed sampling method which combining over sampling and under sampling to overcome the problem about imbalance data, then excavate out improved SMOTE – Tomeklinks-SVM identification model with the best learning ability and generalization ability, and use the model to recognize profit manipulation of China's listed companies. Finally, the empirical results show that under the condition that the sample profit manipulation is far less than the normal sample, the traditional SVM model cannot effectively identify profit manipulation, but after introducing the hybrid sampling method which combining the over sampling and under sampling into the traditional SVM can most effectively enhance the identification accuracy to profit manipulation; The concealment of the means of profit manipulation will make the identification model misjudgement, but internal motive indicators can fully explain to the reason of profit manipulation of listed companies and improved generalization ability of SMOTE Tomeklinks- the SVM model, thus combination of external operation indicators and internal motive indicators can more accurately identify the profit manipulation of listed companies; and in the same level of sample and index, the improved SMOTE Tomeklinks- the SVM model has better generalization ability than other artificial intelligence model; in addition, in the four kinds of kernel function, the SMOTE-Tomeklinks-SVM models with linear kernel function and RBF kernel function are relatively have better recognition ability, but in general, four kinds of kernel function has little influence on the recognition of profit manipulation, in the same way, apart from the sample parameters ?, the change of model parameters have no significant effect to recognition results of improved SVM model, that SMOTE- Tomeklinks- SVM model has good stability.Therefore, this paper argues that SMOTE-Tomeklinks-SVM model which comprehensive external operation indicators and internal motive indicators can more effectively identify the profit manipulation of listed companies in China, which can provide valuable decision-making reference for the investment subject and the government management.