Dimensionality Reduction Methodology in Regional Innovation Potential Analysis: A Competitive Approach

This study aims to develop a methodology for determining an effective mathematical toolkit for dimensionality reduction in terms of parameters of a complex property of a system expressed through integral indicators. This involves examining such methods of mathematical data processing as the principal component analysis, fuzzy logic method, and weighted moving average. Special attention is given to defining the innovation potential of a region and compiling a system of indicators characterizing it. The developed methodology consists of two sequential stages. The first stage involves the application of three data processing methods to normalized input data separately, while the second is associated with obtaining resultant integral values or their groups for each method separately. The first stage is related to solving the following tasks: identifying categories of complex property, determining category indicators, normalizing data, and aggregating integral indicators. The second stage involves defining integral indicators, hierarchical clustering, and identifying the current priority method of information processing through the analysis of the quality produced based on data clustering.