Research On The Financial Risk Of The GEM Listing Corporation Based On Random Forest

The financial crisis in 2008 has swept the world, and has had a different degree of impact on the world economy, and its impact on the world is still in a certain range. Financial risk is a kind of signal transmission,and in a certain extent, it reflected the enterprise management objectively. Small and medium enterprises have an absolute advantage in today’s market number, the key force to the rapid economic growth, and provide an important channel for employment, and they are the company type with both active and development prospects. However, they are often relatively small,and they are easy to go bankrupt in the fierce competition of the market because financing is very difficult due to the constraints of various aspects, and they are faced with the greater financial risk than those large enterprises.Our country launched the gem in the Shenzhen stock exchange on October 23rd,2009,and its original intention is to provide a new financing platform for the small and medium scale enterprises, reducing their financing burden to some extent. But these small and medium enterprises are also facing many problems, such as weak foundation, the the market environment of increasingly fierce competition,in addition they are in the growth period after listing,and their management ability is relatively poor, and the situation is unstable, so they firmly face greater financial risks. Therefore, it is necessary to how to evaluate and predict the financial risk of the enterprise objectively and reasonably,and strengthen the control of the financial risk, and to ensure the healthy and stable development of the GEM market.At present, the index system of the financial risk evaluation is not suitable for high growth, high-tech, high risk and high income of the GEM Listing Corporations,but mainly for the main board market. Therefore, according to the actual situation of the GEM market development, the establishment of the evaluation system for the financial risk of the GEM Listing Corporation can provide the information about the financial risk to the stakeholders, and then they can make the targeted decision, in addition, it can also help regulators to build effective regulatory measures.The article’s innovation is mainly reflected in two aspects:one is the criteria of classification in financial risk companies and normal companies, and they are not classified according to the words "ST" before the company’s name; another aspect is reflected in the application of random forest algorithms in classification problem of imbalanced data.This paper regard all the companies listed on the GEM in Shenzhen in the the data of financial aspect by the end of 2014 as the starting point. Through introducing the theory of financial risk evaluation and forecast of domestic and foreign financial review and GEM listed companies and financial risk briefly, this paper selects 22 financial indicators of the financial situation of companies listed in GEM to study based on principles of constructing the index system of principles; then filtering and sorting data, using R software to remove the invalid data and select the valid data sample, the final effective sample companies has 397; Finally, using random forest algorithm to research the financial situation of the sample companies.There are two conclusions through using filtering and random forest algorithm to select features:Firstly, if using random forest method to model, it needn’t too much processing on the data; secondly,the model performance using a selected subset of features is not necessarily better than the model constructed by the original index set,so we should choose the index set as input variables of the model through comparing. The article uses random forest algorithm to handle the classification of the imbalanced data,and measure the noise tolerance, as a result,it shows:We can deal with imbalanced data classification problems by random forest method; the model constructed by the random forest method can tolerate relatively high noise, so random forest algorithm is very suitable for researching financial risk of listed companies.The paper puts forward three suggestions through researching and analising the financial situation of companies listed on GEM:Firstly, companies listed on GEM should establish the culture of financial risk assessment and prediction; Secondly, the company should focus on the capacity of profitability and growth which is reflected by three feature subsets in this paper; finally,using this model to analyze the quality of companies in management can provide more information for market participants for the regulatory agencies.This paper also has some shortcomings:in the selected listing Corporations, the total number of manufacturing companies is more than 70%, indicating a serious imbalance in the industry, but it is not considered in the paper. As one of machine learning methods, random forest regards decision tree as the base classifier, which can introduce in some of other classification algorithms such as K-nearest neighbor, supporting vector machine and other combination classification algorithms and then make a comparision of the performance and applicability of models of these algorithms,so algorithm of combination classification is one of the important research directions in the future.