Using Distributed Query Result Caching to Evaluate Queries for Parallel Data Mining Algorithms

An increase in the speed of data mining algorithms can be achieved by improving the efciency of the underlying technologies. Query engines are key components in many knowledge discovery systems and the appropriate use of query engines can impact the performance of data mining algorithms. By taking advantage of hypothesis generation patterns, queries, generated from the hypotheses, can be evaluated more e ciently. Caching query results and using the cached results to evaluate new queries with similar constraints reduces the complexity of query evaluation and improves the performance of data mining algorithms. In a multi-processor environment, distributing the query result caches can improve the performance of parallel query evaluations. This idea has been used in the ParDRI system and has resulted in signi cant improvements in the execution times of ParDRI.