An Empirical Study Of Performance Forecasting Method On Listed Enterprises Assets Restructuring

With the development of Chinese stock market, the listed companies are facing anomaly financial situation or other abnormal situations, asset restructuring has become one of the important methods which many listed companies use to improve the performance. It will be the most effective and vitality way adopted by Chinese listed companies. The trend shows the advantages of asset restructuring that it can help ST companies adjust their industrial structure, optimize resource allocation, improve business structure and enhance the asset operational performance, which makes ST companies break away from the title successfully. The forecast of asset restructuring performance is the method to judge whether ST companies could transform qualitatively, and plays an important role in management. Therefore, forecasting asset restructuring performance of ST company and improve the forecast accuracy is of great urgency.Firstly, based on the listed companies’asset restructuring, the data sets show the imbalance of successful asset restructuring and failure of restructuring. Although this obeys the objective reality, relatively compared with the traditional machine learning algorithms, it often brings results to most of the class. Thus, for minority class, it will get the forecast effect relatively poor. In order to improve the accuracy, this paper balances the gathering data and adopts a single forecast model to predict whether asset restructuring could bring enterprises great performance. Ten methods are chosen for performance comparison, including logit、probit、SVM、MDA、CBR、 BaggingLOGIT、BaggingSVM、BaggingPROBIT、BaggingCBR and BaggingMDA. The results show the accuracy of SVM and CBR is better than other eight prediction models. That means more than80%of the ST companies know whether they can recover after asset restructuring by using these models. This research provides basis for improving the performance of ST companies through asset restructuring.Secondly, in order to improve the forecast accuracy, this article uses clustering algorithms with single forecast models and clustering ensemble algorithms with single forecast models, which is distinguishable to the previous researches. Ten prediction models are built, namely CLOGIT, CPROBIT, CMDA, CSVM, CCBR, BaggingCLOGIT, BaggingCPROBIT, BaggingCMDA, BaggingCSVM and BaggingCCBR, to predict the asset restructuring performance and compare function of the ten models. The results show that the model built with clustering algorithms and clustering ensemble algorithms has a positive performance no matter in the accuracy or TPR and TNR. For the accuracy and TPR, clustering ensemble model is superior to clustering model; while on the respect of TNR, clustering model is better than clustering ensemble model.Thirdly, in order to validate the above20models, this paper collects ST companies between2011and2012as the new sample sets to test these models’importance and practical value. The experimental results show that prediction performance of traditional statistical model is inferior to that of artificial intelligence model. The predictive results of the support vector machine and its integration, clustering formation model are relatively stable and effective in the imbalance or balance data; while the performance of case-based reasoning and its integration, clustering formation model is relatively better in imbalance data.Finally, this paper summarizes the research and points out the related management revelation. It plays an important role in the practical application, and provides certain theoretical basis for decision making of supervisors.