Abstract
A mathematical technique for macroscopic social systems modeling is proposed. The main consideration is devoted to detecting structural changes from actual data. The proposals are a simultaneous analysis of data classification and regression, and a fuzzy modeling technique, in the course of which the structural differences between regions are detected and modeled convincingly. A concrete example presented is related to the world population. With a relatively new data covering the world, a fuzzy model consisting of three rules is developed and applied to time series data of Japan to show the effectiveness of the model when used for long-term prediction whereas the used data are obtained in only short period. The paper also discusses some issues on development of time-series models, detection of non-linear substructures, selection of adequate explanatory variables, and prediction of uncertain future.
