Performance study of Association Rule Mining Algorithms for Dyeing Processing System

Abstract

In data mining, association rule mining is a popular and well researched area for discovering interesting relations between variables in large databases. In this paper, we compare the performance of association rule mining algorithms, which describes the different issues of mining process.  A distinct feature of these algorithms is that it has a very limited and precisely predictable main memory cost and runs very quickly in memory-based settings. Moreover, it can be scaled up to very large databases using database partitioning. When the data set becomes dense, (conditional) FP-trees can be constructed dynamically as part of the mining process.  These association rule mining algorithms were implemented using Weka Library with Java language. The database used in the development of processes contains series of transactions or event logs belonging to a dyeing unit. This paper contributes to analyze the coloring process of dyeing unit using association rule mining algorithms using frequent patterns.  These frequent patterns have a confidence for different treatments of the dyeing process.  These confidences help the dyeing unit expert called dyer to predict better combination or association of treatments.  Therefore, this article also proposes to implement association rule mining algorithms to the dyeing process of dyeing unit, which may have a major impact on the coloring process of dyeing industry to process their colors effectively without any dyeing problems, such as shading, dark spots on the colored yarn and etc. This article shows that LinkRuleMiner (LRM) has an excellent performance for various kinds of data to create frequent patterns, outperforms currently available algorithms in dyeing processing systems, and is highly scalable to mining large databases.  This paper shows that HMine and LRM has an excellent performance for various kinds of data, outperforms currently available algorithms in different settings, and is highly scalable to mining large databases. These studies have major impact on the future development of efficient and scalable data mining methods.

Keywords: Performance, predictable, main memory, large databases, partitioning, Weka Library