| Summary: | 碩士 === 義守大學 === 資訊工程學系 === 91 === The objective of this research is to study how to improve the efficiency of the discovery of Adaptive-Support Association Rules (ASAR) for collaborative recommendation systems. Collaborative recommendation (sometimes known as collaborative filtering) is a process by which information on the preferences and actions of a group of users is tracked by a system which then, based on the patterns it observes, tries to make useful recommendations to individual users. Many data mining techniques have recently been proposed for the construction of collaborative recommendation systems, in particular, the fixed step-size adjustment adaptive-support association rule algorithm in [7].
In this work, we propose two adjustable step-size data-mining algorithms to discover the adaptive-support association rules from transaction databases, namely Bisection-based ASAR algorithm and Secant-based ASAR algorithm. Experimental comparisons with the fixed step-size adjustment approach show that our proposed techniques require less computation, both running time and iteration steps, and will always find a corresponding minimum support for the association rules.
|
|---|
