Corporate failure prediction problems are of great interest to researchers as well as creditors, and other interested parties that are likely to rely on corporate financial statements (David et al., 1997). According to Duffy (1999), a large number of major public and private companies have gone failed over the last few years. The recent spectacular failures of such large corporations as Parmalat in Europe, Enron and WorldCom in USA and HIH in Australia are such reminders of the reality: the corporate failures are still ongoing. This study is aimed to develop and validate corporate failure forecasting models using various methodologies. The purpose of these proposed models is to include: (a) assisting banker and investors to intervene in good time to try to prevent company failure; and (b) acting as decision support system for senior management to improve company performance. By reviewing previous studies and researches, totally, 9 types of modelling techniques were adopted in this research, including (i) five types Artificial Neural Networks (ANNs) , which are Multi-layer Perceptron Neural Network (MLP-NN), Learning Vector Quantization (LVQ), Radial Basis Function Neural (RBFN), Self-Organization Mapping (SOM) and Probabilistic Neural Network (PNN), (ii) two statistical methods, which are Multivariate Discriminant Analysis (MDA), Logit analysis and (iii) two hybrid systems, Neuro-Fuzzy and Neuro-Genetic models. 2 separate datasets were collected from with certain sampling criterion, dataset I selected from FAME with 50 samples of UK manufacturing companies, while dataset 11 was collected from AMESDUS with 100 samples of European manufacturing companies. In total, 58 models built-up through the 4-stage experiment plan. According to experiment results of 58 developed models, the stand alone MLP-BP classifier shows critical superior prediction ability than other models, with 97.19 % training accuracy and 97.53% testing accuracy, the overall classification accuracy is 100%, in other words, the ANN stand-alone classifier can predict correctly of healthy companies and failed companies from manufacturing industry.
| Identifer | oai:union.ndltd.org:bl.uk/oai:ethos.bl.uk:685938 |
| Date | January 2008 |
| Creators | Ying, Zhou |
| Publisher | University of Manchester |
| Source Sets | Ethos UK |
| Detected Language | English |
| Type | Electronic Thesis or Dissertation |
Page generated in 0.0017 seconds