Application of Stationary Wavelet Support Vector Machines for the Prediction of UK Economic Recession and Expansion Periods

Abstract

This paper examines the efficiency of various approaches on the classification and prediction of economic expansion and recession periods in United Kingdom. Four approaches are applied. The first is discrete choice models using Logit and Probit regressions, while the second approach is a Markov Switching Regime (MSR) Model with Time-Varying Transition Probabilities. The third approach refers on Support Vector Machines (SVM), while the fourth approach proposed in this study is a Stationary Wavelet SVM modelling. The findings show that SW-SVM and MSR present the best forecasting performance, in the out-of sample period. In addition, the forecasts for period 2012-2015 are provided using all approaches.