利用非線性模式對並存的短期時間序列做科技替代預測

Abstract

此論文利用非線性成長曲線模式對並存的短期時間序列資料做科技替代預 測.其中共變異 數矩陣Σ為AR(1)相關結構.當應用冪次轉換於模式時,我 們介紹兩種非線性建模方法.另一 方面我們利用冷卻模凝方法解決最佳 化問題.此外我們藉由實際資料在非線性模式與(DBT) 模式間做一些預測 精確性的比較. In this paper we use nonlinear growth curve models for forecasting technol ogical substitutions with concurrent short time series data when the covarianc e matrix Σ is a AR(1) covariance structure. While applying power transformati ons, two methods of nonlinear modeling are investigated. Meanwhile, simulate d annealing for optimization problem is also studied. Some comparisons in pred ictiveaccuracy between data-based transformed and nonlinear models are also ma de viareal data.