Applying parallel programming and high performance computing to speed up data mining processing

Water quality assessment and prediction of Lake Michigan are becoming major challenges in Northwest Indiana, USA. Traditionally, mechanistic simulation models are employed for water quality modeling and prediction. However, given the complicate nature of Lake Michigan in Northwestern Indiana, the detailed simulation model is extremely simple in comparison and, at some point, additional detail exceeds our ability to simulate and predict with reasonable error levels. In this regard, my project applied data mining technologies, as an innovative alternative, to develop an easy and more accurate approach for water quality assessment and prediction. The drawback of the data mining modeling is that the execution takes quite long time, especially when we employ a better accuracy but more time consuming algorithm in clustering. Therefore, we applied the High Performance Computing System of the Northwest Indiana Computational Grid to deal with this problem. Up to now, the pilot experiments have achieved very promising preliminary results. The visualized water quality assessment and prediction obtained from this project would be published in an interactive website so that the public and the environmental managers could use the information for their decision making.