Improving Classification Accuracy of IC Packaging Products Database Based on Variable Precision Rough Sets
Titel:
Improving Classification Accuracy of IC Packaging Products Database Based on Variable Precision Rough Sets
Auteur:
Yung-Hsiang Hung
Verschenen in:
Information technology journal
Paginering:
Jaargang 7 (2008) nr. 3 pagina's 440-449
Jaar:
2008
Inhoud:
The aim of the study is to present the Variable Precision Rough Sets (VPRS) methodology to improving the classification accuracy for IC Packaging Product Database. The selection of proper packaging and manufacturing procedures is one of the priorities in IC design operations. Separation of industrial production characteristics require that IC designers have to ask the IC packaging engineers down the supply chain to confirm information regarding related product information, such as IC packaging type, size, functional features and price before they can select IC packaging products and determine IC product design procedures. In response to these product information demands from IC designers, service staffs of IC packaging factories integrate information from various departments to provide feedback to the designers. However, as the related information is very complex and professional, precious time is wasted during their communications, which instigates a failure to meet the demand of quick response. Furthermore, the IC product design and development cost and timing may be affected by lack of information. In an age of IC packaging technological breakthroughs, IC packaging structure and technological capabilities have already become key stages of IC design and manufacturing processes. Hence, it has become a very significant topic of the design industry, to effectively and rapidly, obtain ample product related information of IC packaging to meet their operational demands and enable them to improve operational costs and shorten processes. The present study adopts the VPRS method, an improved rough set theory to be applied to classification accuracy of the IC packaging product database; then compare the highest accuracy values with the values and rules obtained from the Johnson rough algorithm. The experimental results prove that proper β values, based on VPRS, are able to improve IC packaging product classification accuracy to obtain more consistent values and simpler rules regarding maximum value.
Uitgever:
Asian Network for Scientific Information (provided by DOAJ)
Tijdschrift beschrijving