Research On The Behavior Of Large Shareholders Selling Lifted Shares Based On BP Neural Network

Before the split share structure reform, stock market shares can be divided into tradable shares and non-tradable shares by whether or not being traded in the Stock Exchange. This state of equity division resulted in interests of non-tradable shareholders and tradable shareholders distorted to the securities market being unable to play the functions of the resource optimization. With the split share reform in full swing, the limit of tradable shares and non-tradable will gradually be eliminated, and China's securities market will usher in full circulation times. If major shareholders have a substantial reduction in lifted shares, and the market have not enough capability to undertake, then the result of market expansion may lead to excessive hemorrhage and collapsed. Therefore, the analysis of impact of the lifting restrictions on large shareholders to sell shares reduction factors and reduction of major shareholders'motivation to find more effective ways to diminish its negative effects; under the existing fair conditions, disclose financial statements of listed companies to build the stable and reasonable prediction model correctly predicts that the reduction degree of large - scale non-circulation shareholders by reasonability selecting impact indications of reduction of reduction of large - scale non-circulation shareholder. These have great practical significance for the protection of the interests of small investors and government regulators supervising market risk.In this paper, we take the large - scale non-circulation in Shenzhen stock market as the sample. There are 14 factors selected as independent variables from the corporate ownership structure, the performance of listed companies ,own situation of the shareholder and other aspects, and take the proportion of underweight lifted shares of major shareholders as the dependent variable for empirical analysis. This article first through the creation of BP neural network model, analyze the affecting factors of its behavior base on contribution rate of BP neural network. We get from the study that the six factors such as the total stock proportion by the top ten shareholders, ROE, is the top shareholder or not and is the state-owned shares or not and so on play an important role in affecting large shareholders'behavior. We can get contrary correlation curve of the subtraction of shareholder and the ROE if given the same binding conditions, different control level of the shareholders and different Shareholders property. The results of empirical research show that large - scale non-circulation selling shares is not representative of the declining investment value of the enterprise, ordinary investors can get positive investment signals from it. In addition, regulators should completely guarantee the circulation right of large - scale non-circulation to maintain its confidence in the national policy and control the behavior of large - scale non-circulation based on those factors with higher contribution rate, which is more effective to maintain the market stability than limiting it only . Finally, applying BP neural network and rough set theory to building a prediction system for subtraction degree of large - scale non-circulation. The test results show that the system has a high average accuracy and high practicality, which can be provided a reference for ordinary investors and regulators. The study method has reusability and the conclusion provide us theoretical guidance to keep the healthy development of capital markets to full circulation market.