Search

We present an end-to-end interpretable deep architecture that predicts the success of drug prescription based on multiple graph kernel fusion using a graphical representation of electronic health records. We formulate the predictive model as a binary graph classification problem with a set of graph kernels proposed to capture different aspects of graph structures through deep neural networks. Results using the Taiwanese National Health Insurance Research Database demonstrate that our approach outperforms current start-of-the-art models on accuracy and interpretability. The approach is in preliminary deployment.

Predictive models for drug prescription exist; we propose an additional such model that uses a graphical representation of an electronic health record with the prescription process formulated as a kernelized binary (success or failure) classification problem. Our approach improves accuracy while maintaining a high level of interpretability. Results using the Taiwanese National Health Insurance Research Database (NHIRD) favorably compare our approach to other models.

The sports market has grown rapidly over the last several decades. Sports outcomes prediction is an attractive sports analytic challenge as it provides useful information for operations in the sports market. In this study, a hybrid basketball game outcomes prediction scheme is developed for predicting the final score of the National Basketball Association (NBA) games by integrating five data mining techniques, including extreme learning machine, multivariate adaptive regression splines, k-nearest neighbors, eXtreme gradient boosting (XGBoost), and stochastic gradient boosting. Designed features are generated by merging different game-lags information from fundamental basketball statistics and used in the proposed scheme. This study collected data from all the games of the NBA 2018–2019 seasons. There are 30 teams in the NBA and each team play 82 games per season. A total of 2460 NBA game data points were collected. Empirical results illustrated that the proposed hybrid basketball game prediction scheme achieves high prediction performance and identifies suitable game-lag information and relevant game features (statistics). Our findings suggested that a two-stage XGBoost model using four pieces of game-lags information achieves the best prediction performance among all competing models. The six designed features, including averaged defensive rebounds, averaged two-point field goal percentage, averaged free throw percentage, averaged offensive rebounds, averaged assists, and averaged three-point field goal attempts, from four game-lags have a greater effect on the prediction of final scores of NBA games than other game-lags. The findings of this study provide relevant insights and guidance for other team or individual sports outcomes prediction research.

Recognition of unnatural control chart patterns (CCPs) is an important issue because the unnatural CCPs can be associated with specific assignable causes negatively affecting the manufacturing process. By assuming that an unnatural CCP is a combination of normal pattern and process disturbance, a multi-stage control chart pattern recognition scheme which integrates independent component analysis (ICA) and support vector machine (SVM) is proposed in this study. The proposed multi-stage ICA-SVM scheme first uses ICA to extract independent components (ICs) from the monitoring process data containing CCPs. The normal pattern and process disturbance hidden in the process data can be discovered in the ICs. Then, the IC representing the process disturbance can be identified. Finally, the identified IC and the data of the monitoring process are used as input variables to develop three different SVM models for CCP recognition. The simulation results show that the proposed multi-stage ICA-SVM scheme not only produces accurate and stable recognition results but also has better classification accuracy than four competing models.

Artificial neural networks (ANNs) have been found to be useful for sales/demand forecasting. However, one of the main shortcomings of ANNs is their inability to identify important forecasting variables. This study uses multivariate adaptive regression splines (MARS), a nonlinear and non-parametric regression methodology, to construct sales forecasting models for computer wholesalers. Through the outstanding variable screening ability of MARS, important sales forecasting variables for computer wholesalers can be obtained to enable them to make better sales management decisions. Two sets of real sales data collected from Taiwanese computer wholesalers are used to evaluate the performance of MARS. The experimental results show that the MARS model outperforms backpropagation neural networks, a support vector machine, a cerebellar model articulation controller neural network, an extreme learning machine, an ARIMA model, a multivariate linear regression model, and four two-stage forecasting schemes across various performance criteria. Moreover, the MARS forecasting results provide useful information about the relationships between the forecasting variables selected and sales amounts through the basis functions, important predictor variables, and the MARS prediction function obtained, and hence they have important implications for the implementation of appropriate sales decisions or strategies.

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.

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.

The objective of the proposed study is to explore the performance of credit scoring using a two-stage hybrid modeling procedure with artificial neural networks and multivariate adaptive regression splines (MARS). The rationale under the analyses is firstly to use MARS in building the credit scoring model, the obtained significant variables are then served as the input nodes of the neural networks model. To demonstrate the effectiveness and feasibility of the proposed modeling procedure, credit scoring tasks are performed on one bank housing loan dataset using cross-validation approach. As the results reveal, the proposed hybrid approach outperforms the results using discriminant analysis, logistic regression, artificial neural networks and MARS and hence provides an alternative in handling credit scoring tasks.

Data mining is a very popular technique and has been widely applied in different areas these days. The artificial neural network has become a very popular alternative in prediction and classification tasks due to its associated memory characteristics and generalization capability. However, the relative importance of potential input variables and the long training process have often been criticized and hence limited its application in handling classification problems. The objective of the proposed study is to explore the performance of data classification by integrating artificial neural networks with the multivariate adaptive regression splines (MARS) approach. The rationale under the analyses is firstly to use MARS in modeling the classification problem, then the obtained significant variables are used as the input variables of the designed neural networks model. To demonstrate the inclusion of the obtained important variables from MARS would improve the classification accuracy of the networks, diagnostic tasks are performed on one fine needle aspiration cytology breast cancer data set. As the results reveal, the proposed integrated approach outperforms the results using discriminant analysis, artificial neural networks and multivariate adaptive regression splines and hence provides an efficient alternative in handling breast cancer diagnostic problems.

Since solely using statistical process control (SPC) and engineering process control (EPC) cannot optimally control the manufacturing process, lots of studies have been devoted to the integrated use of SPC and EPC. The majority of these studies have reported that the integrated approach has better performance than that by using only SPC or EPC. Almost all these studies have assumed that the assignable causes of process disturbance can be identified and removed by SPC techniques. However, these techniques are typically time-consuming and thus make the search hard to implement in practice. In this paper, the EPC and neural network scheme were integrated in identifying the assignable causes of the underlying disturbance. For finding the appropriate setup of the networks' parameters, such as the number of hidden nodes and the suitable input variables, the all-possible-regression selection procedure is applied. For comparison, two SPC charts, Shewhart and cumulative sum (Cusum) charts were also developed for the same data sets. As the results reveal, the proposed approaches outperform the other methods and the shift of disturbance can be identified successfully.

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 network's 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.

Nowadays, the capability to both generate and collect data has been expanded enormously and provides us with huge amount of data. Millions of databases have been used in business data management, scientific and engineering data management, and many other applications. It is noted that the number of such databases keeps growing rapidly because of the availability of powerful and affordable database management systems. To compete effectively in such an environment, business managers must take advantage of high-return opportunities in time. But however, the difficulty of discerning the value of information keeps many companies from capitalizing fully on the wealth of data at their disposal. Data mining is the art of finding patterns in data and is a new approach based on a general recognition that there is undraped value in large databases and utilities data-driven extraction of information. However, it is still not easy to identify the complicate relationship in the huge data set. Moreover, in most...

Sales forecasting of computer products is regarded as an important but difficult task since computer products are characterized by product variety, rapid specification changes and rapid price declines. Artificial neural networks (ANNs) have been found to be useful techniques for sales forecasting. However the inability to identify important forecasting variables is one of the main shortcomings of ANNs. For selecting an appropriate number of forecasting variables which can best improve the performance of the neural network prediction model, a commonly discussed data mining technique, multivariate adaptive regression and splines (MARS), is adapted in this study. The proposed model, firstly, uses the MARS to select important forecasting variables. The obtained significant variables are then served as the inputs for two neural network models-support vector regression (SVR) and cerebellar model articulation controller neural network (CMACNN). A real sales data collected from a Taiwanese computer dealer is used as an illustrative example. Experimental results showed that the obtained important variables from MARS can improve the forecasting performance of the SVR and CMACNN models. The proposed two-stage forecasting models provide good alternatives for sales forecasting of computer products.

In this study, a hybrid model using multivariate adaptive regression splines (MARS) and SVR is proposed for sales forecasting of information technology (IT) products. Support vector regression (SVR) has become a promising alternative for forecasting due to its generalization capability in obtaining a unique solution. However, one of the key problems is that SVR can not identify which forecasting variables are more important for building the forecasting model. For selecting an appropriate number of forecasting variables which can best improve the performance of the prediction model, a commonly discussed data mining technique, multivariate adaptive regression and splines (MARS), is adapted in this study. The proposed model first uses the MARS to select important forecasting variables. The obtained significant variables are then served as the inputs for the SVR model. Experimental results from three IT product sales data reveal that the obtained important variables from MARS can improve the forecasting performance of the SVR models. The proposed hybrid model outperforms the results using single SVR and single MARS models and hence provides an efficient alternative for IT product sales forecasting.

Credit scoring and behavioral scoring have become very important credit risk management tasks during the past few years due to the impact of several financial crises. The objective of the proposed study is to explore the performance of behavioral scoring using three commonly discussed data mining techniques-linear discriminant analysis (LDA), backpropagation neural networks (BPN), and support vector machine (SVM). To demonstrate the effectiveness of behavioral scoring using the above-mentioned techniques, behavioral scoring tasks are performed on one bank credit card dataset in Taiwan. As the results reveal, BPN outperforms other techniques in terms of overall scoring accuracy and hence is an efficient alternative in implementing behavioral scoring tasks.

Background: Surgery is the only long-lasting effective treatment for morbid obesity. Prediction on successful weight loss after surgery by data mining technologies is lacking. We analyze the available information during the initial evaluation of patients referred to bariatric surgery by data mining methods for predictors of successful weight loss. Methods: 249 patients undergoing laparoscopic mini-gastric bypass (LMGB) or adjustable gastric banding (LAGB) were enrolled. Logistic Regression and Artificial Neural Network (ANN) technologies were used to predict weight loss. Overall classification capability of the designed diagnostic models was evaluated by the misclassification costs. Results: We studied 249 patients consisting of 72 men and 177 women over 2 years. Mean age was 33±9 years. 208 (83.5%) patients had successful weight reduction while 41 (16.5%) did not. Logistic Regression revealed that the type of operation had a significant prediction effect (P=0.000). Patients receiving LMGB had a better weight loss than those receiving LAGB (78.54%±26.87 vs 43.65%±26.08). ANN provided the same predicted factor on the type of operation but it further proposed that HbA1c and triglyceride were associated with success. HbA1c is lower in the successful than failed group (5.81±1.06 vs 6.05±1.49;P=NS), and triglyceride in the successful group is higher than in the failed group (171.29±112.62 vs 144.07±89.90;P=NS). Conclusion: Artificial neural network is a better modeling technique and the overall predictive accuracy is higher on the basis of multiple variables related to laboratory tests. LMGB, high preoperative triglyceride level, and low HbA1c level can predict successful weight reduction at 2 years.