ماهالانوبیس - سیستم تاگوچی - الگوریتم شبکه های عصبی برای داده کاوی در محیط های پویا
|کد مقاله||سال انتشار||مقاله انگلیسی||ترجمه فارسی||تعداد کلمات|
|22148||2009||6 صفحه PDF||سفارش دهید||3412 کلمه|
Publisher : Elsevier - Science Direct (الزویر - ساینس دایرکت)
Journal : Expert Systems with Applications, Volume 36, Issue 3, Part 1, April 2009, Pages 5475–5480
Data-mining analysis has two important processes: searching for patterns and model construction. From previous works finding that the Mahalanobis–Taguchi System (MTS) algorithm is successful and effective for data-mining. Conventional research in searching for patterns and modeling in data-mining is typically in a static state. Studies using a dynamic environment for data-mining are scarce. The artificial neural network (ANN) algorithm can solve dynamic condition problems. This study integrates the MTS and ANN algorithm to create the novel (MTS–ANN) algorithm that solves the pattern-recognition problems and can be applied to construct a model for manufacturing inspection in dynamic environments. From the results of the experiment, we find that the methodology of the MTS algorithm can easily solves pattern-recognition problems, and is computationally efficient as well as the ANN algorithm is a simple and efficient procedure for constructing a model of a dynamic system. The MTS–ANN algorithm is good at pattern-recognition and model construction of dynamic systems.
Large collections of data are potential loads of valuable information. So, in data-mining, search and extraction can be difficult and exhaustive processes (Keim & Kriegel, 1994). In other words, data-mining analysis has two important processes: searching for patterns and model construction. Data-mining is a search for relationships and global patterns that exist in large databases, but are ‘hidden’ in vast amounts of data. Therefore, data-mining is an analytical process that explores data in search of consistent patterns and/or systematic relationships between attributes, and then validates findings by applying the detected patterns to new subsets of a system. In the search for suitable patterns, in a previous work, A genetic algorithm (GA) and Fuzzy methods based to construct association rules (Kaya & Alhajj, 2005), A linear correlation discovering (LCD) method based for pattern recognition (Chiang, Huang, & Lim, 2005), and Authors used a three-stage online association rule to mine context information and information patterns (Wang, Tseng, & Hong, 2006). Authors adopted a neural network to generate automated trend analysis of proteomics data (Nicholson, 2006). Simultaneously, estimation of works has been discussed in terms of system lifecycle evaluation and estimation attributes (Daskalaki, Kopanas, Goudara, & Avouris, 2003). Additionally, the classification and clustering domains, such as visualization, web data search, position clustering, and graphs classification, have been extensively discussed (Chang and Ding, 2005, Coenen and Leng, 2007, Das and Datta, 2007 and Nasraoui et al., 2006). In the field of model construction, a GA algorithm based to model a bankruptcy prediction model (Kim & Han, 2003), an artificial neural network based to predict subsidence (Ambrozic & Turk, 2003), and the use of biblio-mining frameworks to generate a usage-based forecasting rule (Nicholson, 2006). Obviously, when searching for patterns, the MTS is a good and an effective algorithm (Das & Datta, 2007). The MTS developed by Taguchi is a novel method that combines the Mahalanobis distance (MD), orthogonal arrays (OA) and the signal-to-noise (SN) ratio. The MTS is a diagnostic and forecasting method. The main aim of the MTS is to make accurate predictions in multidimensional attributes by constructing a global measure meter. The application of MTS in the pattern-recognition area is as follows: a MTS based to resolve classification problems ( Das & Datta, 2007). Literature review findings demonstrate that the MTS algorithm is successful and effective for data-mining. Conventional research in searching for patterns and modeling in data-mining is typically in a static state. From above mentions, studies using a dynamic environment for data-mining are scarce. The artificial neural network (ANN) algorithm to solve dynamic and multi-response condition problems. This study integrates the MTS and ANN algorithm as a novel method for dynamic environments, that is, this work adopts the Backpropagation neural network (BPN) owing to its ability to map the complex relationship between input data and corresponding outputs (Keim & Kriegel, 1994). In summary, the MTS–ANN algorithm is utilized to construct a pattern and model a dynamic system. A case study of an electronics company is used to verify and validate the proposed methodology.
نتیجه گیری انگلیسی
In above analysis, the forms of two groups of MD are similar to each other, and lists in Fig. 6 and Fig. 7. Consequently, forming the new pattern from second group, which shows the number value of parameter is changed from 30 to 6, and the MD shape is still has a likely appearance between the two groups. From above mentions, we know that the proposed algorithm is successful applied in pattern forming.In modeling a dynamic system aspect, the ANN algorithm shows us the 6–4–6 structure is the optimal selection from other’s architecture, which shows RMSE is convergence in 0.04. From the results, the methodology of the MTS algorithm can easily solves pattern-recognition problems, and is computationally efficient as well as the ANN algorithm is a simple and efficient procedure for constructing a model of a dynamic system. The MTS–ANN algorithm is good at pattern-recognition and model construction of dynamic systems. We conclude that the MTS–ANN algorithm can successfully be applied to dynamic environments for data-mining problems.