Abstract:
Every second the amount of data in and around us is increasing and there's no end to the sight. Data Mining is done to ease the understanding of large amounts of data by discovering interesting exceptions or regularities. Association rules are simple but powerful regularities in binary data. The problem in Association Rule is that there can be easily hundreds or even more association rules holding in a data set which presents a problem in their utility and understanding. Therefore, Data Mining itself can produce such great amounts of data that there is new knowledge management problem.
The current World realizes the need for compact knowledge discovery in databases and have attempted solutions. I present `A Core Based Approach on Large Item Sets' that efficiently summarizes the information present in the data set. Therefore the concept of conditional entropy to measure information content in a rule is used. A Core is a set that is found in more than one set.
The Compact knowledge discovery can be identified by compactness factors like: degree of useful information conveyed, a measure of strength of a rule, suitable number of rules, measure of interdependence between items and so on.
In this thesis I end up showing a proposed approach that meets some compactness factors and promise to satisfy the
