A Novel Approach for Enhancing Direct Hashing and Pruning for Association Rule Mining

Abstract

Data Mining has been considered as a promising field in the intersection of databases, artificial intelligence and machine learning. Association rule mining has been one of the most popular data mining subjects, which can be simply defined as finding interesting rules from large collections of data. This paper introduces an enhanced hashing approach in discovering associations for large itemsets. The proposed hashing approach scans the entire database only once using the improved version of Apriori Algorithm termed as Direct Hashing and Pruning (DHP) algorithm. The algorithm determines the frequency of each k-itemset and discovers set of rules from frequent k-itemsets. The application uses the minimum support provided by the domain expert and frequent itemsets are discovered. Later the numbers of k-itemsets are reduced in the pruning phase after scanning the database completely. The proposed approach is not prone to collisions yielding high accuracy.

Keywords: Association rule mining, direct hashing and pruning, frequent pattern tree mining