Adjustable discovery of adaptive-support association rules for collaborative recommendation systems

In this work, we propose an adjustable stepsize data-mining algorithm to discover adaptive-support association rules (ASAR) [Lin, B. et al., Jan. 2002] from data sets. Adaptive support association rules are constrained association rules with application to collaborative recommendation systems. To discover association rules for recommendation systems, minimum conference and a specific target item in association rules are usually assumed and no minimum support is specified in advance. Based on size monotonicity of association rules, i.e., the number of association rules decreases when the minimum support increases, an efficient algorithm using adjustable step size for finding minimum support and therefore adaptive-support association rules is presented. Experimental comparison with the fixed step size iterative approach shows that our proposed technique requires less computation, both running time and iteration steps, and would always find a corresponding minimum support.