A Hybrid of EEMD and LSSVM-PSO model for Tourist Demand Forecasting

Abstract

In this research, hybrid model of Least Square Support Vector Machine (LSSVM) and Ensemble Empirical Mode Decomposition (EEMD) are presented to forecast tourism demand in Malaysia. Foremost, the original series of tourism arrivals data was separated using EEMD technique into residual and Intrinsic Mode Functions (IMFs) components. Next, both of IMFs and residual components were forecasted using Particle Swarm Optimization (LSSVM–PSO) method. In the end, the predicted result of IMFs and residual components from LSSVM–PSO method are sum together to produce the forecasted value for tourism arrivals in Malaysia. Empirical results showed that the presented model in this paper outperform individual forecasting model. The result indicated that LSSVM–PSO is a promising tool in time series forecasting by having the presence of non-stationary and non-linearity in the time series data.