Frequent Set Meta Mining: Towards Multi-Agent Data Mining

In this paper we describe the concept of Meta ARM in the context of its objectives and challenges and go on to describe and analyse a number of potential solutions. Meta ARM is defined as the process of combining the results of a number of individually obtained Associate Rule Mining (ARM) operations to produce a composite result. The typical scenario where this is desirable is in multi-agent data mining where individual agents wish to preserve the security and privacy of their raw data but are prepared to share data mining results. Four Meta ARM algorithms are described: a Brute Force approach, an Apriori approach and two hybrid techniques. A “bench mark” system is also described to allow for appropriate comparison. A complete analysis of the algorithms is included that considers the effect of: the number of data sources, the number of records in the data sets and the number of attributes represented.

• Type: Conference
• Year: 2007
• Where: SGAI
• Authors: Kamal Ali Albashiri, Frans Coenen, Robert Sanderson, Paul H. Leng
• Keywords: Data Mining, Meta ARM, Meta ARM Algorithms, SGAI 2007