Searching for an optimal feature subset in a high-dimensional feature space is an NP-complete problem; hence, traditional optimization algorithms are inefficient when solving large-scale feature selection problems. Therefore, meta-heuristic algorithm

Cutting is an essential process in touch panel manufacturing, which concerns the effectiveness of whole touch panel manufacturing process and reliability of final application products. An alternative is laser cutting technology, which is expected to eclipse conventional cutting method in terms of cost and quality. The laser cutting quality of touch panel is determined by the appropriation of value settings of several parameters/factors involved in the complex laser cutting process. However, the parameter/factor adjusting method applied in practice mainly relies on the engineer's experience and trial-and-error experiment, which are unsystematic and susceptible to inefficiency and ineffectiveness. This study proposes a procedure for parameter/factor optimization of the touch panel laser cutting process to obtain maximum cutting quality. The proposed optimization procedure is successfully applied in a real case of laser cutting for projected capacitive touch panel, a type of touch panels using capacitive technology; it gradually becomes mainstream in the newly emerged information appliance marketplace. Results reveal that defect rate of the laser cutting process decreases from 32.6% to 0.3% once the proposed procedure on this paper is implemented, and this achievement outperforms other optimization strategies.

Multiclass Mahalanobis-Taguchi system (MMTS), the extension of MTS, is developed for simultaneous multiclass classification and feature selection. In MMTS, the multiclass measurement scale is constructed by establishing an individual Mahalanobis space for each class. To increase the validity of the measurement scale, the Gram-Schmidt process is performed to mutually orthogonalize the features and eliminate the multicollinearity. The important features are identified using the orthogonal arrays and the signal-to-noise ratio, and are then used to construct a reduced model measurement scale. The contribution of each important feature to classification is also derived according to the effect gain to develop a weighted Mahalanobis distance which is finally used as the distance metric for the classification of MMTS. Using the reduced model measurement scale, an unknown example will be classified into the class with minimum weighted Mahalanobis distance considering only the important features. For evaluating the effectiveness of MMTS, a numerical experiment is implemented, and the results show that MMTS outperforms other well-known algorithms not only on classification accuracy but also on feature selection efficiency. Finally, a real case about gestational diabetes mellitus is studied, and the results indicate the practicality of MMTS in real-world applications.

Under the highly developed automation today, the manufacture of saxophone is still a nonautomatic process and much relies on highly skilled technicians. In order to insure the timbre quality, the sound of finished saxophone must be tested in the final inspection stage. The evaluation of timbre quality mainly depends on the professional musicians' hearing judgment; however, the sensitivity of human perception can be influenced by many factors. To improve the reliability of saxophone timbre quality inspection, an automatic multiclass timbre classification system (AMTCS) is developed and used to assist in the inspection work. The AMTCS is composed of our proposed waveform-shape-based feature extraction method in parameterization phase and multiclass Mahalanobis-Taguchi system in classification phase. The numerical experiments show that the musical instrument classification accuracy obtained by our proposed AMTCS is satisfactory. Through employing the AMTCS, strong assistance was provided to the inspection of saxophone timbre quality, and a perfect identification rate on the saxophones with different timbre quality levels is achieved. Moreover, the significant tones having impact on saxophone timbre quality can also be easily identified by AMTCS.

In classification problems, the class imbalance problem will cause a bias on the training of classifiers and will result in the lower sensitivity of detecting the minority class examples. The Mahalanobis-Taguchi System (MTS) is a diagnostic and forecasting technique for multivariate data. MTS establishes a classifier by constructing a continuous measurement scale rather than directly learning from the training set. Therefore, it is expected that the construction of an MTS model will not be influenced by data distribution, and this property is helpful to overcome the class imbalance problem. To verify the robustness of MTS for imbalanced data, this study compares MTS with several popular classification techniques. The results indicate that MTS is the most robust technique to deal with the classification problem on imbalanced data. In addition, this study develops a "probabilistic thresholding method" to determine the classification threshold for MTS, and it obtains a good performance. Finally, MTS is employed to analyze the radio frequency (RF) inspection process of mobile phone manufacturing. The data collected from the RF inspection process is typically an imbalanced type. Implementation results show that the inspection attributes are significantly reduced and that the RF inspection process can also maintain high inspection accuracy.

Low-temperature polysilicon (LTPS) technology has drawn the attention of many display manufacturers because it has several potential advantages over amorphous thin-film-transistor liquid crystal display TFT LCD). The mapping operation matches one TFT and one color-filter (CF) glass plate together to form one LCD plate. Each plate contains a certain number of cells that are independent devices. The matched LCD cell is good only when the TFT and CF cell match is good. When only a TFT or CF cell is good, there is a yield loss. The sorter is a robot used in LCD manufacturing systems to achieve higher yield in matching TFT and CF plates. This sorter contains several ports that can transfer CF glasses from CF cassettes to match TFT glasses. This is an important determinant for post-mapping yield. This study first proposes a linear programming formulation to optimize the plates-matching problem for the various ports. Next, we propose an algorithm to reduce the number of ways for choosing different matched objects when the number of matched cassettes is large. This algorithm avoids computer overload and provides an excellent solution. The empirical results illustrate the efficiency and effectiveness of the proposed approaches. Note to Practitioners - Yield is an important competitiveness determinant for thin-film-transistor liquid crystal display (TFT LCD) factories. TFT-LCD contains three major manufacturing sectors: the array, cell and module assembly processes. The yield loss from the cell process is one of the most critical steps. This study first formulates a linear programming for the TFT color-filter plates-matching problem in the cell process and a reduction algorithm then is developed to reduce the computational efforts. Our proposed approach is workable and it does not require a significant investment to produce yield improvement by applying the proposed matching algorithm.

The variable precision rough sets (VPRS) model is a powerful tool for data mining, as it has been widely applied to acquire knowledge. Despite its diverse applications in many domains, the VPRS model unfortunately cannot be applied to real-world classification tasks involving continuous attributes. This requires a discretization method to preprocess the data. Discretization is an effective technique to deal with continuous attributes for data mining, especially for the classification problem. The modified Chi2 algorithm is one of the modifications to the Chi2 algorithm, replacing the inconsistency check in the Chi2 algorithm by using the quality of approximation, coined from the rough sets theory (RST), in which it takes into account the effect of degrees of freedom. However, the classification with a controlled degree of uncertainty, or a misclassification error, is outside the realm of RST. This algorithm also ignores the effect of variance in the two merged intervals. In this study, we propose a new algorithm, named the extended Chi2 algorithm, to overcome these two drawbacks. By running the software of See5, our proposed algorithm possesses a better performance than the original and modified Chi2 algorithms.

The study presents a novel semiparametric prediction system for the Taiwan unemployment rate series. The prediction method incorporated into the system consists of a neural network model that estimates the trend, as well as a Box-Jenkins prediction of the residual series. The response surface methodology is employed to find the appropriate setup of network parameters as the neural network is applied. Also, extensive studies are performed on the robustness of the built network model using different specified censoring strategies. In terms of the adaptability of the Box-Jenkins method, the prediction intervals of the system can be successfully constructed. To demonstrate the effectiveness of our proposed method, the monthly unemployment rate from June 1983 to February 1992 is evaluated using a neural network model with Box-Jenkins technique and other alternative methods, e.g. space-time series analysis, univariate ARIMA model and state space model. Analysis results demonstrate that the proposed method outperforms other statistical methodologies.