This project is aimed at developing a methodology for the visualization of multidimensional datathrough an approach based on the use of methods of reduction of dimension and models of human-computer interaction. In particular, the study is oriented to explore the interactive combinationof dimensionality reduction (DR) methods in order to expand the embedded spaces that can beobtained from a database of input, with the aim of giving the user the possibility to choose therepresentation of the data that better meet their needs. During the development of this researchproject, several interaction methods are proposed and two forms of combination of RD methodswere taken into account, the first one combining the resulting embedded spaces and the second onethrough the use of kernel approaches. Finally, based on the results obtained, the methodology isapplied to the creation of a data visualization tool, which incorporates: i) interaction models, ii)a mixture of RD methods, iii) traditional visualization techniques (scatter diagrams and parallelcoordinate diagram). Additionally, with the purpose of generating a dynamic interaction (changesin real time), the algorithm of locally linear sub-matrices is implemented to carry out the dimensionreduction process at lower computational cost. It is important to highlight that the whole tool isdeveloped under scalability and modularity criteria, so that future works (improvements) can beeasily incorporated into it.
This project is intended to build a structured database of respiratory-disease diagnosed patients via spirometry data. Also, as another main goal thereof, a comparative study on pattern recognition techniques applied on spirometry records will be performed, which is aimed at finding those techniques achieving a good compromise between effectiveness, computational cost, and interpretability of the physiological concept.