Of recent, data science methods have been used to study and forecast financial and economic problems. This paper uses historical data to build a more parsimonious predictive model for making short term forecasts of the future values for the Composite Indicator of Economic Activity (CIEA) in Ghana. Based on our studies of a variety of shrinkage methods and a dimension reduction technique, we show empirically that the estimated model based on the Adaptive Elastic Net (Adaptive ENET) algorithm offers the greatest forecasting potential for the CIEA. A major finding in this paper was that, the Adaptive ENET model outperformed the benchmark model: Principal Component Regression (PCR) according to the cross validation root mean square error difference Statistic.
Mathematics Subject Classification: G12; C15; G22
Keywords: Composite Index of Economic Activity; Least Absolute Shrinkage and Selection Operator; Elastic Net; Principal Component; Artificial Neural Network
[ Download ]
