A Conceptual Approach for a Soft Computing Framework to Determine Correlations in High-Dimensional Data

Abstract

AbstractBased on previous results from the “Efficient Airport 2030” cluster of excellence project, we will present a conceptual approach for a Soft Computing Framework which is focused on analyzing high-dimensional data for possible correlations. The data to analyze consists of various (in some cases fuzzy) input parameters that are expected by Web Service based simulation models; the delivered results are taken into account as well. The input can be considered to be an element of the Cartesian product of the sets containing the respective input parameters possible values. Even though the data can be generated by using the simulation models, it is not viable (due to the complexity) to compute the data as a whole. Being capable of working on fuzzy or incomplete data, Soft Computing provides an interesting approach that we expect to lead to a suitable solution to find correlations without the need to compute the data for all possible inputs.The presented framework is separated into several parts: We will present a concept for an abstraction layer for the generation of the data to gain independence from the actual data source. We will further provide the needed data structures to be able to represent several input parameters with arbitrary types as a tuple. Furthermore we will provide a possibility to use user-defined (complex) data types. We will present an initial selection of methods we intend to use, alongside with their integration in the framework. We will also give a first idea on how to use Soft Computing methods to find correlations in (high dimensional) data.