Abstract

In recent years, many data systems have added data mining components. Some have developed data mining tools as a key component. These tools query the database using ad hoc queries. Often, ad hoc queries fall victim to performance problems. These performance problems occur because the database is not optimized for the queries. At design time data correlation has been identified as a means of improving the performance of queries in data systems. This paper investigates two methods that allow database administrators to enhance the performance of data systems when ad hoc query are used. They are data correlation and an adaptive software agent.

The central method discussed in this paper expands on the idea that data correlation can be used to enhance the performance of databases. Our technique uses point access methods and correlation statistics to create an index schema that attempts to enhance the performance of ad hoc queries in business and scientific data.

An adaptive software agent based on unsupervised and reinforcement learning algorithms was also developed and is shown to follow a user query pattern over time. This software agent adapts to the way users query a relation. The software agent is shown to adapt to the user's needs over time enhancing the performance of ad hoc queries.