کشف دانش با استفاده از رویکرد عصبی جهت تجزیه و تحلیل مسئله ریسک اعتباری شرکت های کوچک و متوسط در ترکیه
|کد مقاله||سال انتشار||تعداد صفحات مقاله انگلیسی||ترجمه فارسی|
|18689||2011||6 صفحه PDF||سفارش دهید|
نسخه انگلیسی مقاله همین الان قابل دانلود است.
هزینه ترجمه مقاله بر اساس تعداد کلمات مقاله انگلیسی محاسبه می شود.
این مقاله تقریباً شامل 4440 کلمه می باشد.
هزینه ترجمه مقاله توسط مترجمان با تجربه، طبق جدول زیر محاسبه می شود:
Publisher : Elsevier - Science Direct (الزویر - ساینس دایرکت)
Journal : Expert Systems with Applications, Volume 38, Issue 8, August 2011, Pages 9313–9318
This study proposes a knowledge discovery method that uses multilayer perceptron (MLP) based neural rule extraction (NRE) approach for credit risk analysis (CRA) of real-life small and medium enterprises (SMEs) in Turkey. A feature selection and extraction stage is followed by neural classification that produces accurate rule sets. In the first stage, the feature selection is achieved by decision tree (DT), recursive feature extraction with support vector machines (RFE-SVM) methods and the feature extraction is performed by factor analysis (FA), principal component analysis (PCA) methods. It is observed that the RFE-SVM approach gave the best result in terms of classification accuracy and minimal input dimension. Among various classifiers k-NN, MLP and SVM are compared in classification experiments. Then, the Continuous/Discrete Rule Extractor via Decision Tree Induction (CRED) algorithm is used to extract rules from the hidden units of a MLP for knowledge discovery. Here, the MLP makes a decision for customers as being “good” or “bad” and reveals the rules obtained at the final decision. In the experiments, Turkish SME database has 512 samples. The proposed approach validates the claim that is a viable alternative to other methods for knowledge discovery.
Credit risk analysis (CRA) recently attracts more attention since credit volume in real market has shown great increase and economical fluctuations has become more often. Credit risk (CR) is a general term which implies to future losses. CRA aims to decrease future losses by estimating the potential risk and eliminating the new credit proposal if the risk is higher than a defined tolerance value. This is also called as CR classification which labels a customer as “good” if he could pay the loan back, otherwise as “bad”. In Turkey, 95% of real enterprises are accepted as small and medium enterprises (SME) that reveals the importance of them in national economy (Fantazzini & Figini, 2009). Not only in Turkey, but also in many developing countries in the world, especially in recent years, SME credits have been gaining much more importance according to their high growth in financial world. In contrast with its increasing growth rate in the world-wide financial sector, there is a limited research for SMEs’ CRA. In our study, we highlight the potential of neural networks as tools of knowledge discovery in SMEs CRA problem in Turkey. A trained multilayer perceptron (MLP) makes a credit decision as being “good” or “bad” of customers and our neural rule extraction algorithm discovers the knowledge embedded in the MLP. First, various input attribute selection methods are used to select a minimal input dimension. Although, small subset of the original SME portfolio is used as an input, when the real, huge portfolio size is taken into consideration; the dimension reduction becomes indispensable phase. Decision Tree (DT) and Recursive Feature Elimination with Support Vector Machine (SVM-RFE) are applied for feature selection, and then Factor Analysis (FA) and Principal Component Analysis (PCA) are applied for feature extraction. In the experiments, six permutations of this data set were obtained and they were labeled as Dataset1, Dataset2, … , Dataset6. Then, the purpose of the study is to show that for the six datasets, it is possible to achieve good accuracies. More importantly, this can be achieved by neural network with one hidden layer and having enough connections which makes it possible to extract comprehensible rules using our rule extraction algorithm, Continuous/Discrete Rule Extractor via Decision Tree Induction (CRED). We first present the results from applying MLP, SVM and k-NN classifiers on the six datasets. After the good accuracy rates were achieved, we address the issue of knowledge discovery from the neural networks. For this purpose, we obtain rules from the CRED algorithm using trained MLP with both continuous and discrete features. The work flow of our proposed method is given in Fig. 1.
نتیجه گیری انگلیسی
In this research, we focus on the knowledge discovery for CRA which becomes very important in real financial market in the recent years. We proposed a CRED based knowledge discovery method for CRA of Turkish SME customer portfolio, which covers both customer classification and rule-base extraction. From the experimental results with limited amount of data, it is observed that the proposed model well-performs on the sample space however as stated earlier, we do not affirm that the model would well-perform on the whole portfolio. We also state that several important rules were extracted for our application. The obtained results cannot be compared to any other free dataset because SMEs behavior changes according to many different conditions from country’s financial regulations to the economical fluctuations of data collection duration. Our results have to be supported by the larger and balanced datasets in Turkish SME customer portfolio in order to discover a general knowledge in the area.