Classification of operating performance of the enterprises is not only a hot issue emphasized by the management, but it is even the important reference by investors in their decision-making. In general, the analysis of its performance is usually undertaken by models of financial prediction or credit rating. This paper address a lot of models to analyze it through the financial ratio from 287 private enterprises of traditional industry public listed in Taiwan's stock market and OTC as sample data. A hybrid methodology that combines both data mining and artificial intelligence is proposed to take advantage of the unique strength of single one model. First, we use the data mining technique, such as traditional principal components analysis, to select network input variables. Second, the various different models, including the Probabilistic Neural Network are also considered. Third, this paper shows that the classification ability of the Probabilistic Neural Network model, after the parameter adjusted by genetic algorithm, does significantly outperform other simple methods-back-propagation network, decision tree, and logistic regression model. In conclusion, experimental results with real data sets indicate that combined model can be an effective way to improve forecasting classification accuracy achieved by either of the one single models.
Since Particle Swarm Optimization (PSO) has properties such as: fast convergence, the ability to search global optimum and very strong universal characteritistic, it is thus very suitable to be used in clustering analysis and parameter utilization of optimized neural network by the researchers. Therefore, in this article, it is used to applied in analyzing enterprise's Financial Characteristic. First, in this article, based on the profit force and growth force of financial five forces, the financial ratio data of companies with stocks listed in regular and over-the-counter stock market in Taiwan and in financial crisis are collected, meanwhile, two normal enterprises with similar characteristics are collected for pairing purpose. Furthermore, with the aim of deriving profit force and growth force, respectively, Grey Relational Analysis is done; in the mean time, the analytical results of both of them are ranked according to grey relational grade so as to understand the performance ranking of each enterprise in profit force and growth force; then PSO is used to divide it into two groups, and the financial characteristics of these two groups of companies are compared, and the results can be used as reference by managers in the enterprises; finally in this article, three data mining techniques such as: PSO Grey Model Neural Network, Genetic Algorithm Optimized Grey Model Neural Network and general Grey Model Neural Network are used, respectively to set up Enterprise Financial Distress model and Enterprise Financial Characteristic detection model. The anlysis indicates that two different groups can be divided based on PSO. One group is enterprises that excel in profit force and growth force while the other group is enterprises that are not good at both of them. On the other hand, in Enterprise Financial Distress model and Enterprise Financial Characteristic model, the PSO Grey Model Neural Network model demonstrates the fastest convergence and the best classification capability.
Clustering analysis is the basis of the construction of many classifications and systems with main objective to plan and divide the data into many subsets according to certain rules. Due to the practical function of clustering analysis, many researchers thus proposed different clustering algorithms to be used by researchers around the world. In this article, Fuzzy Sammon Mapping method is associated to perform clustering effect and classification capability analysis on these frequently used clustering algorithms. From the result of test data of an investigation of banking service satisfaction, it was found that GK Cluster algorithm can have very good clustering effect; however, for classification capability, hard clustering analysis method is better.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.