The Study Of Character And Identification Of Special-Treated Company Based On Particle Swarm Optimization Algorithm

Facing to the current global economy which is from prosperity to crisis and to recovery now, financial markets go through an unprecedented unrest. Many companies that run well come to naught, and some work to survive with difficulty. Why would the public company be so vulnerable which had better days in the past? Do they have the well-operating conditions, as the gorgeous "surface" of the company? Many problems exposed by the financial crisis, and led to suspect the real operating conditions of the public companies, and to lose the confidence to the public companies by people. How the investors invest and the government control the economic marketing has become huge barriers which affect the development of financial markets regularly.Furthermore, the environment of the financial-marketing is complicated and diversified. The investors speculate the real state of the public company with difficulties, and just because of this, the investment risk enhanced. Many young investors have often become the genuine sacrificial victims when the public companies went bankrupt. At last the young investors are compelled to withdraw from the capital market, and then it would impact on the size of capital market and economic development, and would lead to another economic crisis.From the above mentioned, we can find the necessity to identify the operating conditions of the public companies. The identification could guide the investors to the correct direction, and to avoid the risks of the investment, what is more, from the national perspective, it also could supervise the stock-marketing and contain the opportunism. There are long-term studies in the analysis of operating conditions of the enterprise especially the financial aspect. Its importance has been well known to the academia. Along with the multiplex tendency of the world economics, Operating conditions of enterprises have also become "multi-up". Therefore, it is even more important to arrive at the indicator-system and model which can identify the operating conditions of the enterprise effetely. Its high value nature and applications become the Focus and hot spots of academics who have studied at home and abroad. Regarding the domestic research, the public companies which were "ST" had already been plunged into the crisis of management. So, we can focus on the "ST" tab at beginning of the research, and then propose a model to identify and provide early warnings for the investors as a basis of investment analysis and decision-making. The model will identify whether the good statement of the public companies or not, and forecast them whether can be pasted on "ST" label or not. This is this research main idea. The research has also provided the feasibility for the recognition of public companies'statement.The paper mainly focuses on the financial reports and notes which are provided by the public companies. The author will also put the data into the "vacuum" when forecasting, and does not take other factors into account, such as human factors, environmental factors, speculative factors, and so on. For the choice of the data,the author extract the real ones to avoid the affecting of the fraud ones.At the beginning of the paper, it mainly introduces the characteristics which could reflect the statement of the companies. There are a total of forty-six characteristics from the four perspectives which are profitability, solvency, efficiency and income, covering as much as possible to reflect the status of public companies in all aspects of business-operating. This step builds the foundation for the model based on Particle Swarm Optimization and feature selection which would do after. Particle Swarm Optimization (PSO) is one of the Swarm Intelligence systems, which is a "bionic" algorithm that imitates the group-behaviors of birds' foraging. In the research, it is found that the birds could enhance the efficiency of food's searching through sharing the location of the food, when they looking for food. In the practical problems, we may consider the result as a "bird", that "the particle". Each "particle" also has its own speed and location information. "Particles" could continuously move through the exchange of the information. By repeated exchange-iteration, we will find the "final food". PSO is an effective global optimization algorithm, which implement easily, have the highly precision of searching and fast convergence. For the advantages, the academia has paid more attentions to the PSO and has widely used in a variety of optimization problems in practice.Cluster analysis technology is a process that divides the date into multiple clusters which are composed of similar objects according to the differences between the data objects and the particular judge rules. What the paper studies is distinguishing companies between ST and non-ST, and determine even to predict the attribution of the samples according to the distance between the sample and final cluster centers. The commonly used cluster algorithms have the K-MEANS cluster algorithm, the ISODATA algorithm, the revision ISODATA algorithm and so on. The algorithm used in this paper is K-MEANS cluster algorithm which has the advantages of fast convergence and good results, however, has the disadvantages of the tendency to fall into local optimal solution. The paper combines Particle Swarm Optimization Algorithm and K-MEANS clustering algorithm to build models which could make up the disadvantages of the K-MEANS.According to the problem the paper should solve, besides the model, it needs the character indexes which can reflect the actual problem. There are some characters which do not have any relations to the problem in the data mining and will affect accuracy and efficiency of the data mining. So before the application of data analysis, the author will analyze all the character of the operating conditions of the public companies firstly, and filter out the ones which have the greatest impact, finally construct the subset of the characters used to the model. There are two common algorithms which is Filter and Wrapper to make the feature selection. What the paper used is Filter. According to the Filter, author will get the sequence of characters, and put the characters which have the most important impact on the statement of the public companies into the PSO model, eventually arrive at the conclusion of the paper.