Research On Financial Failure Prediction Based On Two-Level DEA

With the accelerating pace of informationization and globalization of the economic, market competition is increasingly fierce, and there has radical change in global economy, our country enterprise especially listed companies accept the unprecedented severe challenges. Under such hash economic and competitive environment, enterprise is more easy to fall into financial difficulties its survival is becoming more difficult. Because the financial condition of a company is not only affect its own survival and development but affect the interests of the investors, creditors and other interested parties, corporate financial failure prediction is playing an increasingly important role.Since the 1930s, there are various financial failure prediction models in the related literature, and researchers try to improve these models continuously to strive to build a better early warning model. Data Envelopment Analysis (DEA) was introduced to determine the relative efficiency of a set of similar Decision Making Units (DMU), where each DMU uses multiple inputs to produce a number of outputs. In this paper, we will establish predicting model based on the two-level DEA model and "worst practice frontier" (WPF) DEA, and with companies listed in China’s Shenzhen and Shanghai Stock Exchange as samples for the empirical research. The main research contents of this paper include the following:Firstly, structure the two-level DEA predicting model. The discrimination power of DEA will be weakened if too many input or output indicators are used. Especially, when the number of DMUs is relatively small, most of these DMUs will be evaluated to be efficient. It is well accepted that many output and input indicators which are similar or highly corrected are needed to comprehensive evaluation of corporate financial condition by reasearchers. Two-level DEA approach can deal with many indicators in the same time by arranging the inputs and outputs in two hierarchy levels by categorizing them into separate groups according to their financial similarity. To construct early warning model with two level indicators system based on the two-level DEA model, and use financial data of three years of 75 companies listed on China’s Shenzhen A-share for empirical validation.Secondly, structure the WPF two-level DEA predicting model. The WPF-DEA model picks out financial failure companies based on how bad they perform in the worst-case condition. Beacause the companies under the poor-case environment are easy to fall into financial failure, so we should find out the companies of potential financial failure in the poor-case environment effectively in financial failure prediction. First of all, structure WPF two-level DEA model combining the two-level DEA and WPF DEA. Then to propose the predicting model based on the two level indicators system. Lastly, to select the companies listed on China’s Shenzhen and Shanghai Stock Exchange from 2012 to 2013 years as samples for the empirical research.Thirdly, select indicators and determine the weights. This study selects indicators those have high correlation with the company’s financial situation and separate them into four categories according to their financial similarity to establish two level indicators system. Then, use a scientific method combining the SVM and normalization method to determine the weights of indicators.Establish effective accurate financial failure predicting models through the scientific method is not only greatly enriched the financial failure prediction theory, also has a lot of practical significance. The predicting models help company managers detect the company’s financial problems timely, investors invest scientifically, and creditors manage the loans and receivable accounts payable, etc reasonably. Besides, financial failure prediction provides the management basis for the government supervision departments, at the same time enrich its regulatory means to maintain the financial market order and safeguard social stability.