Research On The Risk Assessment Model For Listed Companies On The GEM Based On ASVM

The opening of the Growth Enterprise Market (GEM) has important practical and strategicsignificance for alleviating problems of financing difficulties facing small-medium enterprises(SME), promoting sustainable economic development and improving the internationalcompetitiveness. However, GEM listed companies have characteristics of immaturity, dynamic,future uncertainty and so on, so the risk is even more concerned by the market. There are very fewliteratures on China’s Growth Enterprise Market to study the risk assessment. The establishment of ascientific risk assessment model for listed companies on the GEM is very important for venturecapitalists, the listed company on GEM and the investors in the secondary market to identify andpredict risks accurately and to improve risk management capability.Within the context of the listed companies on GEM, this dissertation investigates the riskassessment methods. Our aim is to help the interested parties in GEM to assess business riskaccurately and to improve the risk control level. In this dissertation, firstly, we establish a riskassessment index system for the listed companies on GEM. Secondly, we propose a risk assessmentmodel based on support vector machine approach using historical data and the imbalance of risk dataon GEM and the decision-makers’ risk preferences are considered in the proposed model. Thirdly,with the aid of the experts’ domain knowledge and experience, we establish coordination frameworkfor groups in which the corresponding cost constraints for reaching acceptable consistency isconsidered. In order to obtain more effective results, we establish an evaluation model that makesfull use of the advantages of subjective decision-making model and of data mining models.Therefore, the subjective and objective information are integrated. Finally, the risk assessmentmodel based specific–attribute risk is constructed with some dynamic features. Detailed contents ofthis dissertation are as follows.1. In this dissertation, we first summarize and review the risk factors affect venturesystematically. Then, we conclude the preliminary risk assessment index system using the balancedscorecard thoughts from five perspectives: financial, marketing and customer, internal processes,management and employees as well as environment. When designing risk assessment index system,we also consider the differences between companies listed on the GEM and companies listed on theMain Board in the industry distribution, the company size, income structure, in which the life-cyclestage, corporate governance structure and stability. In this dissertation, the risks of companies listedon the GEM are reflected by both financial information and non-financial information. After thepreliminary risk assessment index system is obtained, weights of all indexes are given through usingquestionnaire, experts’ advice and the final risk index can be selected by combining with the ratingof importance of index and the availability of the index.2. The risk assessment model for the companies listed on the GEM is induced from historicaldata using data mining techniques. An adaptive support vector machine is proposed for thecharacteristics of risk data in the domestic GEM such as small samples of data, poor information,nonlinear, unbalanced and other complex features listed companies. Different preferences of thedifferent decision-makers for the risk of misclassification are also considered. For the riskassessment model based on SVM is not strong interpretability under the nonlinear case, weights ofall indexes are extracted from SVM through using the quadratic programming techniques.3. In this dissertation, we build risk assessment models for companies listed on GEM throughusing experts’ methods within the framework of multi-attribute group decision making based on AHP. In this dissertation, for improving the consistency of comparison matrix, the consensus of theexperts and the control mechanisms of interaction, a cost constrained group decision makingframework was proposed that includes two feedback mechanisms: the first feedback mechanism is toadjust the experts’ weight according to their contribution to the decision making; the second is toguide the experts to improve the quality of their own judgments based on the opinions andperspectives.For the strategies commonly used in the integration of subjectivity and subjectivity can notreally integrate experts’ subjective judgments into the process of solving the model, there may be adeparturing from experts’ original judgment or objective weight determined by history informationand questions emerging out of subjective adjustment when outcomes of decision-making have beenobtained. On the basis of statistics, regarding experts’ weight as a real sample in the decision-makinggroups, this dissertation obtains the true weights using ASVM within the estimated weight of theweight distribution in the sample interval.4. Considering the risk evolution process of index of assessment, this dissertation structuresASVM based on process information to assess the risk for listed companies on the GEM. Commonlyused risk assessment model for a simple single-structure model of cross-sectional data, trendinginformation on the risks reflect the dynamic evolution of insufficient capacity and not attachingimportance to issues such as subjective judgments of decision makers, the dissertation, on the use ofdifferent characteristics of the indicators, measures the risks in different ways. Single cross-sectionof the data can reflect the state of variables, including expectations of an S-function target metricsinherent risks, time series data reflecting the need to process variables, the use of modern financialtheory of asset risk measure, considering the expectations of policy makers on the indicators, timeseries of the mean, variance or skewed distribution and other characteristics, including informationon trends in enterprise risk evolution of time series data mapped to a cross section. So the riskassessment model has the ability to handle dynamic information.