Query processing in e-commerce environment using predictive partitioned relations

Too many attributes in a relation are not relevant to fulfilling the user's requirement. Different clients may be interested different set of attributes in the relation. Given this situation, how can the Database Management Systems process relevant attributes instead of reading all attributes? In a data-warehousing environment, materialised view techniques are used and the relation can be vertically partitioned into different sets. If there is c number of clients, then c number of relations will be partitioned. In this paper, we discuss data mining techniques that select most relevant attributes in the relation using predictive partitioned relation. The goal of this research is to locate within a relation those areas of attributes containing tuples relevant to fulfilling the user's requirement. These areas can then be given to a human or automated system for extraction of information, thereby saving a user or query processing system from reading or processing the entire attributes.