Search

Latin hypercube sampling is widely used in industrial engineering. In the traditional Latin square sampling, engineers often arrange sample points in the feasible domain uniformly. However, in practice, engineers may have some prior information about the sub-domains where the response volatility is relatively large, named the interesting sub-domains. In order to make full use of these information, this paper employed the D-S evidence theory to fuse prior information from different sources/fields. Then we divide the feasible domain into different sub-domains and indicate the interesting sub-domains. For the sample placement, we put more points in these interesting sub-domains and less points in other sub-domains. Finally, we construct the model with the proposed sample points placement approach based on prior information. A case study was conducted to illustrate the proposed method. The case study shows that the proposed method performs better than the traditional model in MSE, MaxE and StdE.

Understanding how certain media types affect consumers' preference for foreign brands is a major concern for managers of Multinational Corporations to make better media choices in advertising campaigns. However, little research has been conducted to facilitate managers' decision. A rough set based approach to predict consumers' preference for foreign and domestic brands in their purchasing decisions is proposed. Using data collected from a large scale survey, a case study is conducted to illustrate the application and advantages of the proposed approach.

Firms need to find customers who are both willing and able to participate in creative activities. Therefore, using the customer response model and value model separately or only using the value model is inadequate to identify the qualified customers. A model which simultaneously estimating the response model and the value model is proposed to identify customerspsila participation membership. The empirical study using data from a commercial bank showed that the proposed model is better than the response model and the value model for firms to identify the qualified participative customers.

Encouraging customerspsila repurchase behavior is becoming one of the most important goals for many firms. However, existing methods have their limitations in accurately predicting repurchase behavior. They either require independence and normality assumptions of predicting variables, or have the danger of over-fitting, or result in poor generalization performance. A support vector machines (SVM) based method is proposed to predict customerspsila repurchase behavior. After using sequential pattern to discover repurchase behavior, SVM was used to classify and predict repurchase behavior. The empirical study using customerspsila data from a commercial bank shows that, SVM doesnpsilat require specific assumption of variables; the prediction error of the proposed method decreases by 37% and 54% respectively compared with those of logistic regression and artificial neural network; moreover, both the prediction error and its standard deviation decrease with the increase of sample size. Those evidences demonstrate the effectiveness and superiority of the proposed method.

Based on the extensive literature review and the philosophy of integration of quality tools, the paper develops integrated quality information system (IQIS) software, and applies it to a manufacturing company. The software integrates the quality tools as well as the process quality information. It provides guidance for locating bottleneck process through integrated data analysis and also supports Six Sigma process improvement. The result shows that the application of IQIS can optimize the process of design and manufacturing, shorten the cycle time of product, reduce the cost, and realize quality improvement continuously.

Dual response surface methodology (DRSM) and nonparametric methodology (NPM) are main approaches used to achieve robust parameter design (RPD) of industrial processes and products. When the relationship between influential input factors and output quality characteristic of a process is very complex, both approaches have their limitations. For DRSM, it fails to fit the real response surfaces of process mean and variance by using the second order polynomial models. For NPM, it is hard to optimize parameters of fitting equation, and it needs more experiments as well. From a machine learning perspective, this paper generalizes RPD as a restricted active learning problem and proposes a new approach to achieve it. It fits process mean and variance responses by support vector machines (SVM), and then optimizes levels of design parameters by genetic algorithm. In order to reduce experiment times, the influence of priori knowledge on generalized error of fitting model is studied. Then a prior knowledge based experiment design is developed. Moreover, the approach selects the form of kernel function and optimizes parameters in SVM by comparing the upper bounds of generalized error of different SVM models without extra samples. The example given in the paper shows that, the generalized error and the experiment times of the approach decrease by no less than 45% and 39% respectively, compared with traditional approaches. All these results demonstrate the adaptability and superiority of the approach proposed in the paper.