Application of Data Mining Tools for Exploring Data: Yarn Quality Case Study
No Thumbnail Available
Files
Date
2008-11-24
Authors
Journal Title
Series/Report No.
Journal ISSN
Volume Title
Publisher
Abstract
Businesses are constantly striving for a competitive edge in the economy, and data-driven decision making is crucial to achieve this goal. Four data mining tools, principal component analysis, cluster analysis, recursive partitioning, and discriminant analysis, were used to explore the major factors that contribute to ends down in a rotor spinning manufacturing process. Principal component analysis was used to explore the research question about whether the large number of cotton properties used to classify cotton could be reduced to a significant few. Cluster analysis was used to gain insight about whether there were groups of gins, counties, or classing offices that produced better raw material than others and led to less ends down. The important research question of what raw material properties were affecting ends down was explored with both recursive partitioning and discriminant analysis. Additional research investigated the effect of cotton variety and atmospheric conditions on spinning productivity. Each of the four data mining tools used was informative and offered a different perspective to the overall research question. Several significant factors emerged including humidity, temperature, %DP 555, and uniformity in addition to micronaire and the color properties (+b and Rd). With these results the researcher developed an improvement plan for better control and increased spinning productivity in future operations. A designed experiment is necessary to thoroughly investigate the impact of certain factors beyond the exploratory conclusions obtained from this study.
Description
Keywords
discriminant analysis, recursive partitioning, cluster analysis, principal component analysis, data mining, yarn spinning
Citation
Degree
MS
Discipline
Textile Engineering
