Modeling of nonstationary time-series data. Part I. Data with regular periodic trends

This paper is Part-I of a two-part paper in which composite modeling approach to nonstationary time series forecasting is proposed. Time-series data with regular periodic trends are considered in this paper and data with dynamic trends are treated in the second part of the paper. The major data comprising components are generalized by deterministic models identified through three-step sequential fitting procedure using non-linear regression technique. The proposed modeling approach has been applied to two data sets obtained from two different fields and both containing regular periodic trends.