NSF Award
0232775
This Small Business Innovation Research (SBIR) Phase I project aims on creating a visualization technology for advanced analysis of large multivariate datasets. Such technology will be harnessed for finding hidden dependences in data, revealing its cluster structure, selection of features, and finding regions of extreme values of the functions defined on the data. Successful solutions of such problems require preserving the most important distances between data patterns in a lower dimension. Currently existing<br/>visualization methods only preserve distances for datasets of the size up to few thousands. Many applications in the process industries, bio-informatics, medicine, and defense include analysis of datasets containing large datasets with 10,000-1 million patterns. The hierarchical technique suggested in this project will expand the use of visualization for advanced data analysis to that range of datasets. The first applications<br/>are seen in process industries such as power generation. Broader use of this innovation is anticipated in medicine, bio-informatics, and defense for knowledge discovery through revealing the structure of data. This project will be used in education for teaching multivariate analysis. <br/> The Phase I activities are expected to create a foundation for further development in Phase II, leading to integration in a commercial software package. The first applications of this visualization technology will be for building models for use in coal-fired power plants advanc pollution control systems. The visualization tool will become part of a commercially available package and will be incorporated into a more comprehensive data analysis, modeling, optimization, and control systems. Medicine, bio-informatics, and defense are seen as additional potential beneficiaries of the visualization tool.
