Modified algorithms for synthesizing high-frequency rules from different data sources

摘要

Because of the rapid growth in information and communication technologies, a company’s data may be spread over several continents. For an effective decision-making process, knowledge workers need data, which may be geographically spread in different locations. In such circumstances, multi-database mining plays a major role in the process of extracting knowledge from different data sources. In this paper, we have proposed a new methodology for synthesizing high-frequency rules from different data sources, where data source weight has been calculated on the basis of their transaction population. We have also proposed a new method for calculating global confidence. Our goal in synthesizing local patterns to obtain global patterns is that, the support and confidence of synthesized patterns must be very nearly same if all the databases are integrated and mono-mining has been done. Experiments conducted clearly establish that the proposed method of synthesizing high-frequency rules fairly meets the stipulation.