This Article Full Text (PDF) References Alert me when this article is cited Alert me if a correction is posted Services Email this article to a friend Similar articles in this journal Similar articles in Web of Science Alert me to new issues of the journal Add to Saved Citations Download to citation manager Request Permissions Request Reprints Add to My Marked Citations Citing Articles Citing Articles via Google Scholar Citing Articles via Scopus Google Scholar Articles by Yin, X.-G. Articles by Yu, W. Search for Related Content Social Bookmarking                         What's this?

Selection and Evaluation of Input Parameters of Neural Networks Using Grey Superior Analysis

College of Textiles, Dong Hua University, Shanghai 200051, People's Republic of China, Wuxi Entry-Exit Inspection & Quarantine Bureau, Wuxi 214 101, PR China

Weidong Yu

College of Textiles, Dong Hua University, Shanghai 200051, People's Republic of China, Wuhan University of Science and Engineering, Wuhan 430074, People's Republic of China, wdyu{at}dhu.edu.cn

Grey superior analysis (GS) has been primarily studied in order to select the efficient input variables of an artificial neural network during the worsted yarn manufacturing. The analysis of the processed data indicates that the parameter selection by means of the grey relevancy matrix can obtain the corresponding sequence according to their correlation degree and derive out a group of main factors as the input variables with high grey-correlation for the ANN model. Through the actual calculation and prediction accuracy analysis, the parameters selected by using the grey superior analysis are more correct and effective but less in number than those produced by the subjective and empirical method popularly used in the field of the textile industry. The prediction and the optimization of the processing techniques can be executed with the ANN model optimized by GS. Moreover, three coefficients for the evaluation of the validity of the input variables selected to ANN have been put forward, which are covering experience coefficient, , experience redundant coefficient, β, and new information coefficient, . Comparing the subjective and experiential method (SE) with GS, = 0.30~0.60, β = 1.00~ 2.75, and = 0.17~0.35, that means the traditional used SE method is only partial correct with a relatively high redundancy.

Key Words: grey superior analysis • neural network • parameter selection • optimization

Textile Research Journal, Vol. 77, No. 6, 375-386 (2007)
DOI: 10.1177/0040517507077481


CiteULike    Complore    Connotea    Del.icio.us    Digg    Reddit    Technorati    Twitter    What's this?