Time series modelling and forecasting of Singapore property price: an optimal control approach

Abstract

Examines the formulation of a time series autoregressive model using an optimal control framework. The objective of the model is to forecast the future value of the time series based upon the current and some past values of the time series. In this approach, we cast the parameters of the time series autoregressive model as the optimal control coefficients in order to obtain an improved forecast. This allows effective use of information underlying the system dynamics to drive the evolution of the autoregression parameters. This is in contrast to the typical constant-parameters autoregressive models commonly used for time series modelling and forecasting. The proposed optimal control based time series model is tested on the Singapore property price index. The results show a significant improvement of the forecasting performance over the constant-parameters autoregressive models and the random walk models.