2009•Unpublished venueRequires access

Developing Novel and Effective Approach for Association Rule Mining Using Progressive Sampling

V. Umarani, M. Punithavalli

Open publisher page 8 citations

Abstract

A challenging task in data mining is the process of discovering association rules from a large database. Most of the existing association rule mining algorithms make repeated passes over the entire database to determine the frequent itemsets, which is likely to incur an extremely high I/O overhead. A simple but an effective way to overcome this problem is to sample the database, such that, it produces rules with highest achievable accuracy on the large database. Numerous researchers have proposed sampling approaches for faster and efficient mining of association rules. In this paper, we propose a novel and effective progressive sampling-based approach for mining association rules from a large database. Initially, the frequent patterns are extracted using Apriori algorithm from an initial sample that is selected based on the temporal characteristics and the size of the database. Using the frequent itemsets generated, the negative border of the initial sample is obtained and sorted. Subsequently, the midpoint itemset in the sorted negative border is scanned in the concrete database to check if it is frequent. Based on the support level computed for the midpoint itemset, the sample size is either progressively increased for determining an optimal sample or association rules are mined by considering it as an optimal sample. The experimental results demonstrate the efficiency of the proposed progressive sampling approach in effective mining of association rules.

About this research paper

What this paper is about

A challenging task in data mining is the process of discovering association rules from a large database. Most of the existing association rule mining algorithms make repeated passes over the entire database to determine the frequent itemsets, which is likely to incur an extremely high I/O overhead. A simple but an effective way to overcome this problem is to sample the database, such that, it produces rules with highest achievable accuracy on the large database. Numerous researchers have proposed sampling approaches for faster and efficient mining of association rules. In this paper, we propose a novel and effective progressive sampling-based approach for mining association rules from a large database. Initially, the frequent patterns are extracted using Apriori algorithm from an initial sample that is selected based on the temporal characteristics and the size of the database. Using the frequent itemsets generated, the negative border of the initial sample is obtained and sorted. Subsequently, the midpoint itemset in the sorted negative border is scanned in the concrete database to check if it is frequent. Based on the support level computed for the midpoint itemset, the sample size is either progressively increased for determining an optimal sample or association rules are mined by considering it as an optimal sample. The experimental results demonstrate the efficiency of the proposed progressive sampling approach in effective mining of association rules.

Why it matters

OpenAlex reports 8 citations for this work. Citation counts describe recorded attention and do not establish research quality.

Key contribution

A contribution statement is not available in the OpenAlex record.

Method / approach

Method details are not available in the OpenAlex metadata.

Main findings

Findings are not separately available in the OpenAlex metadata.

Limitations

Limitations are not available in the OpenAlex metadata.

Applications

Application details are not available in the OpenAlex metadata.

Available abstract

A challenging task in data mining is the process of discovering association rules from a large database. Most of the existing association rule mining algorithms make repeated passes over the entire database to determine the frequent itemsets, which is likely to incur an extremely high I/O overhead. A simple but an effective way to overcome this problem is to sample the database, such that, it produces rules with highest achievable accuracy on the large database. Numerous researchers have proposed sampling approaches for faster and efficient mining of association rules. In this paper, we propose a novel and effective progressive sampling-based approach for mining association rules from a large database. Initially, the frequent patterns are extracted using Apriori algorithm from an initial sample that is selected based on the temporal characteristics and the size of the database. Using the frequent itemsets generated, the negative border of the initial sample is obtained and sorted. Subsequently, the midpoint itemset in the sorted negative border is scanned in the concrete database to check if it is frequent. Based on the support level computed for the midpoint itemset, the sample size is either progressively increased for determining an optimal sample or association rules are mined by considering it as an optimal sample. The experimental results demonstrate the efficiency of the proposed progressive sampling approach in effective mining of association rules.

Key concepts: Association rule learning, Data mining, Apriori algorithm, Computer science, Overhead (engineering), Midpoint, Sample (material), Sampling (signal processing)

Related papers

Back to paper searchBrowse research topicsOriginal source
Developing Novel and Effective Approach for Association Rule Mining Using Progressive Sampling — Research Paper | ScholarLens