Article Preview
Top1. Introduction
Few decades before, computer was a simple device used for doing computations, and calculations in a limited area. But emergence of networking and communication technologies, has replaced the role of computer from stand alone system to distributed systems. Simultaneously, the processing speed is considerably increased. This helps in processing data at a greater speed. At present age, enormous amount of data are exchanged, generated, stored, and manipulated through the internet and through numerous sources. But, what is the need of such huge accumulated data unless we extract or predict some useful information from it. So data analysis, information retrieval, and prediction of decisions for unseen associations is of recent research. Additionally, the branch of data mining concerned about the prediction of future probabilities and trends are referred as predictive analysis. It deals with the variables that can be measured based on other single or multiple factors to predict the decision. In traditional approach of predictive modeling, data are collected, a statistical model is formulated and predictions are made with validating the available data. But statistical methods have its own limitations and cannot produce better prediction when the data contains uncertainty. Further to handle uncertainty in predictive data analysis, many intelligent techniques such as rough set (Pawlak, 1982), rough set on fuzzy approximation space (De, 1999), rough set on intuitionistic fuzzy approximation space (Acharjya & Tripathy, 2009), and hybridization of these concepts with other techniques such as neural network, genetic algorithm, formal concept analysis (Tripathy, Acharjya & Ezhilarasi, 2011), etc., were developed. Moreover, predictive analysis is applied in numerous areas such as health sector, telecommunications, financial services, marketing, actuarial science, travel, pharmaceuticals etc. Our basic objective in writing this paper is to make a comparative study between statistical approach and some of the computational intelligent approach. To show the viability of comparison, financial bankruptcy dataset is used to measure the financial distress of the public firms.
Corporate bankruptcy plays a significant role in the field of finance, for the economic phenomena of a country. Policy makers, investors, managers, consumers, industry shares holders, are the prominent entities for the healthy and successful business world (Cielen, Peters, & Vanhoof, 2004). Business failure is a world-wide problem. To enhance the growth throughout the country, some mechanism should be available to predict the number of firms that may fail due to bankruptcy. Simultaneously, the failure serves as an index for the continuous development and robustness of a country's economy (Min & Jeong, 2009; Zhang & Wu, 2011). The consequences raised by the corporate bankruptcies urge the researchers to carry out research work in this direction. Bankruptcy prediction technique is a vast area of finance and accounting research. The research on developing such prediction models initializes its process by focusing on various classification models to distinguish failed and non-failed firms. Such models are of major importance for the budgetary decision makers, as they serve as early-warning system for the failure probability of a corporate entity. To this end, varied traditional statistical methods are employed for predicting financial distress. As stated earlier, in this paper, we compare the statistical approach with various rough computing techniques, using the data collected from the Greek industrial bank, ETEVA, which finances industrial and commercial firms in Greece (Slowinski & Zopounidis, 1995; Greco, Matarazzo, Pappalardo, & Slowinski, 2005). Furthermore, it will help on forming a economic distress prediction system to provide information to the investors, policy makers, and monitoring organizations.
The rest of the paper is organized as follows: Section 2 provides literature review on the bankruptcy prediction models. Section 3, discuss the foundations of the techniques used for predictive data analysis whereas Section 4 explains about the data organization and the proposed research model using various predictive data analysis techniques followed by Section 5, that depicts an experimental comparative analysis, and the paper is concluded by conclusion in Section 6.