Author(s): Anurag Choubey | Ravindra Patel | J.L. Rana
Journal: International Journal of Computer Science & Information Technology
ISSN 0975-4660
Volume: 4;
Issue: 1;
Start page: 221;
Date: 2012;
VIEW PDF
DOWNLOAD PDF
Original page
Keywords: Data mining | Frequent pattern | Graph structure | Adjacency Matrix
ABSTRACT
Association rule mining is a function of data mining research domain and frequent pattern mining is anessential part of it. Most of the previous studies on mining frequent patterns based on an Apriori approach, which required more number of database scans and operations for counting pattern supports in the database. 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 facts 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. Experimental results show the effectiveness of our graphbased approach.
Journal: International Journal of Computer Science & Information Technology
ISSN 0975-4660
Volume: 4;
Issue: 1;
Start page: 221;
Date: 2012;
VIEW PDF


Keywords: Data mining | Frequent pattern | Graph structure | Adjacency Matrix
ABSTRACT
Association rule mining is a function of data mining research domain and frequent pattern mining is anessential part of it. Most of the previous studies on mining frequent patterns based on an Apriori approach, which required more number of database scans and operations for counting pattern supports in the database. 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 facts 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. Experimental results show the effectiveness of our graphbased approach.