Accelerating XML mining using graphic processors

Mining of association rules is an important research direction of data mining. Extensive use of XML on web makes it an interesting source for data extraction from large data sets. There is a growing demand for modern tools and technologies which can efficiently handle such large data. This paper proposes a collaborative approach to extract association rules from structured XML data with the help of cost effective and energy efficient Graphic Processors. The serial approach comprises of deserialization using XPath followed by parallel sorting. In the parallel model there is parallel deserialization of XML data with the help of graphic processor followed by sorting the converted XML data with the help of in-built multithreaded structure of GPU. An empirical performance study on synthetic data is given, demonstrating a remarkable speed increase on a GPU as compared with fully optimized CPU implementation.