Data Mining and Knowledge Engineering

Pradeep Chouksey ¹, Vijay Choudhary ¹, Juhi Singh ¹

Affiliations
1 Technocrats Institute of Technology, Bhopal, IN

Abstract

The discovery of association rules is the most well studied problem in data mining. It has received much attention among the various data mining problems. Many algorithms have been proposed for this purpose. In this paper an efficient algorithm for mining association rules has been proposed which is fundamentally different from known algorithms. Compared to previous algorithms this algorithm reduces the database scans thereby lowering the CPU overhead in most cases. We have performed extensive experiments and compared the performance of this algorithm with that of the best existing algorithm. It was found that the database scans are reduced in the proposed algorithm.

Keywords

Apriori, Association Rules, Frequent Patterns, Transactional Database.

Full Text

     

Pradeep Chouksey ¹, Fiona Jain ²

Affiliations
1 Dept. of M.C.A, Technocrats Institute of Technology, Affiliated to Rajiv Gandhi Technical University, Bhopal, Madhya Pradesh, IN
2 Dept. of MCA, Technocrats Institute of Technology, Bhopal, Madhya Pradesh, IN

Abstract

Exploring association rules, correlations, clusters and many more theories that make out interactions amongst set of items is frequently studied in data mining research, involves complications in mining the data. Mining from the scratch is the most time consuming operation in this discovery process in the computation of the frequency of occurrences of subset of items in the database of transaction, since the size of each set of transaction may be massive that it makes difficult to perform traditional data mining tasks. This research intends to propose a graph structure that captures only those itemsets that needs to define a sufficiently immense dataset into a submatrix representing important weights and does not give any chance to outliers. We have devised a strategy that covers significant facets of data by drilling down the large data into a succinct form of an Adjacency Matrix at different stages of mining process. The graph structure is so designed that it can be easily maintained and the trade off in compressing the large data values is reduced. This proposed research is an explorative way for regular frequent itemsets with fewer scans.

Keywords

Adjacency Matrix, Directed Graph, Multi Level Frequent Patterns, Concept Taxonomy.

Full Text