Mining Queries for Constructing Materialized Views in a Data Warehouse

A data warehouse stores historical information, continuously being generated over time, to support decision making. The queries posed for decision making are usually exploratory, long, complex and analytical in nature. These queries, when posed against a large and continuously growing data warehouse, consume a lot of time for processing and thereby resulting in high response times. This problem of high response time can be addressed by constructing materialized views on the data warehouse. These views, which store data along with its definition, cannot be arbitrarily constructed as they need to contain relevant and required information for answering most future queries. The approach proposed in this paper attempts to identify such information, from previously posed queries on a data warehouse, using clustering and association rule mining techniques. The information identified using the approach is likely to answer most future queries in reduced query response times. As a result, the decision making would become more efficient.