Chih-Chou Chiu

CREDIT SCORING USING THE HYBRID NEURAL DISCRIMINANT TECHNIQUE

Abstract Credit scoring has become a very important task as the credit industry has been experiencing double-digit growth rate during the past few decades. The artificial neural network is becoming a very popular alternative in credit scoring models due to its associated memory characteristic and generalization capability. However, the decision of networks topology, importance of potential input variables and the long training process has often long been criticized and hence limited its application in handling credit scoring problems. The objective of the proposed study is to explore the performance of credit scoring by integrating the backpropagation neural networks with traditional discriminant analysis approach. To demonstrate the inclusion of the credit scoring result from discriminant analysis would simplify the network structure and improve the credit scoring accuracy of the designed neural network model, credit scoring tasks are performed on one bank credit card data set. As the results reveal, the proposed hybrid approach converges much faster than the conventional neural networks model. Moreover, the credit scoring accuracies increase in terms of the proposed methodology and outperform traditional discriminant analysis and logistic regression approaches.

Financial time series forecasting using independent component analysis and support vector regression

As financial time series are inherently noisy and non-stationary, it is regarded as one of the most challenging applications of time series forecasting. Due to the advantages of generalization capability in obtaining a unique solution, support vector regression (SVR) has also been successfully applied in financial time series forecasting. In the modeling of financial time series using SVR, one of the key problems is the inherent high noise. Thus, detecting and removing the noise are important but difficult tasks when building an SVR forecasting model. To alleviate the influence of noise, a two-stage modeling approach using independent component analysis (ICA) and support vector regression is proposed in financial time series forecasting. ICA is a novel statistical signal processing technique that was originally proposed to find the latent source signals from observed mixture signals without having any prior knowledge of the mixing mechanism. The proposed approach first uses ICA to the forecasting variables for generating the independent components (ICs). After identifying and removing the ICs containing the noise, the rest of the ICs are then used to reconstruct the forecasting variables which contain less noise and served as the input variables of the SVR forecasting model. In order to evaluate the performance of the proposed approach, the Nikkei 225 opening index and TAIEX closing index are used as illustrative examples. Experimental results show that the proposed model outperforms the SVR model with non-filtered forecasting variables and a random walk model.

Mining the customer credit using classification and regression tree and multivariate adaptive regression splines

Credit scoring has become a very important task as the credit industry has been experiencing severe competition during the past few years. The artificial neural network is becoming a very popular alternative in credit scoring models due to its associated memory characteristic and generalization capability. However, the relative importance of potential input variables, long training process, and interpretative difficulties have often been criticized and hence limited its application in handling credit scoring problems. The objective of the proposed study is to explore the performance of credit scoring using two commonly discussed data mining techniques-classification and regression tree (CART) and multivariate adaptive regression splines (MARS). To demonstrate the effectiveness of credit scoring using CART and MARS, credit scoring tasks are performed on one bank credit card data set. As the results reveal, CART and MARS outperform traditional discriminant analysis, logistic regression, neural networks, and support vector machine (SVM) approaches in terms of credit scoring accuracy and hence provide efficient alternatives in implementing credit scoring tasks.

Integration of fuzzy theory and ant algorithm for vehicle routing problem with time window

This study intends to apply ant colony optimization algorithm for vehicle routing problem with time window. Fuzzy sets theory is employed to transform time variables into fuzzy variables and then search the shortest vehicle route. The evaluation results showed that setting service level between 70% and 90% has the better result. Besides, the analysis of route and the interval of time window have great impact on the result of route planning.

Data Classification Using the Mahalanobis-Taguchi System

ABSTRACT The Mahalanobis-Taguchi System (MTS) is a pattern information technology developed by Dr. Taguchi. This technology is aimed at providing a better prediction for multivariate data through the construction of a multivariate measurement scale. In this study, two sets of data are analyzed to demonstrate the effectiveness of the MTS. Implementation results reveal that the MTS outperforms traditional discriminant analysis methods. In addition, several important issues regarding the MTS are summarized in the conclusion section.

Neural network forecasting of an opening cash price index

The study investigates the information content of SGX-DT Nikkei 225 futures prices during the non-cash-trading (NCT) period using an artificial neural network model. The cash market closing index, the futures prices from a period in the same trading day and on the following trading day are utilized to determine the appropriate input nodes of a back propagation neural network model in forecasting the opening cash price index. Sensitivity analysis is first employed to address and solve the issue of finding the appropriate network topology. Extensive studies are then performed on the robustness of the constructed network by using different training and testing sample sizes. The effectiveness of the method is demonstrated on data from a 6-month historical record (1998-99). Analytic results demonstrate that the proposed neural network model outperforms a neural network model with the previous days closing index as the input node and the random walk model forecasts. It, therefore, indicates that there is valuable information involved in futures prices during the NCT period that can be used to forecast the opening cash market price index.

Stock index prediction: A comparison of MARS, BPN and SVR in an emerging market

Stock index prediction seems to be a challenging task of the financial time series prediction process especially in emerging markets with their complex and inefficient structures. Multivariate adaptive regression splines (MARS) is a nonlinear and non-parametric regression methodology and has been successfully used in classification tasks. However, there are few applications using MARS in stock index prediction. In this study, we compare the forecasting performance of MARS, backpropagation neural network (BPN), support vector regression (SVR), and multiple linear regression (MLR) models in Shanghai B-Share stock index. Experimental results show that MARS outperforms BPN, SVR and MLR in terms of prediction error and prediction accuracy.

SHIFTS RECOGNITION IN CORRELATED MANUFACTURING PROCESS PARAMETERS USING NEURAL NETWORK

ABSTRACT Because of the existence of the autocorrelation in the data series, traditional statistical process control (SPC) techniques of control charting are not applicable in many process industries. Therefore how to reduce the process variability obtained through the use of SPC techniques has been discussed for years in process industries. Techniques are needed to serve the same functions as SPC control charts that are to identify shifts, in correlated parameters. Neural networks are a massively parallel system and are a potential tool that can be used to identify shifts in correlated process parameters. In this research, backpropagation neural networks are developed to identify shifts in process parameter values from papermaking and viscosity data set available in literature. For finding the appropriate number of input nodes to use in a neural network model, the all-possible-regression selection procedure is applied. For comparison, the time series residual control charts are also developed for the data sets. As the results reveal, networks were successful at identifying data that were shifted one, one and half, and two standard deviations from non-shifted data for both utilized cases. The SPC control charts were not able to classify the same process shifts. In the other words, the neural networks outperform than the traditional SPC approaches in identifying shifts in process parameters. Therefore, it is allowing improved control in manufacturing processes that generate correlated process data.

Forecasting stock price using Nonlinear independent component analysis and support vector regression

In developing a stock price forecasting model, the first step is usually feature extraction. Nonlinear independent component analysis (NLICA) is a novel feature extraction technique to find independent sources given only observed data that are mixtures of the unknown sources, without prior knowledge of the mixing mechanisms. It assumes that the observed mixtures are the nonlinear combination of latent source signals. This study propose a stock price forecasting model which first uses NLICA as preprocessing to extract features from forecasting variables. The features, called independent components (ICs), are served as the inputs of support vector regression (SVR) to build the prediction model. Experimental results on Nikkei 225 closing cash index show that the proposed method can produce the best prediction performance compared to the SVR models that use linear ICA, principal component analysis (PCA) and kernel PCA as feature extraction, and the single SVR model without feature extraction.

Predicting stock index using an integrated model of NLICA, SVR and PSO

Predicting stock index is a major activity of financial firms and private investors. However, stock index prediction is regarded as a challenging task of the prediction problem since the stock market is a complex, evolutionary, and nonlinear dynamic system. In this study, a stock index prediction model by integrating nonlinear independent component analysis (NLICA), support vector regression (SVR) and particle swarm optimization (PSO) is proposed. In the proposed model, first, the NLICA is used as preprocessing to extract features from observed stock index data. The features which can be used to represent underlying/hidden information of the original data are then served as the inputs of SVR to build the stock index prediction model. Finally, PSO is applied to optimize the parameters of the SVR prediction model since the parameters of SVR must be carefully selected in establishing an effective and efficient SVR model. Experimental results on Shanghai Stock Exchange composite (SSEC) closing cash index show that the proposed stock index prediction method is effective and efficient compared to the four comparison models.