Development of pso based hybrid lssvm model for time series prediction

Abstract

Time series forecasting has a fundamental importance in various practical domains. Many models have been proposed in literature to model and predict the Time Series Data (TSD) efficiently. As the modeling and prediction depends on the nature of TSD, one model may not be opt for all applications. This paper presents a hybrid model based on Particle Swarm Optimization (PSO) with Least Square Support Vector Machine (LSSVM). PSO parameter is used for optimizing LSSVM parameters and the proposed model is tested on titan stock market data which is a time series data, Surface roughness data, strength of radiation shielding concrete (RSC) data are non TSD and evaluated using the standard performance measures like Mean absolute error (MAE), Mean Relative Error (MRE) and Root Mean Squared Error (RMSE). It can be observed from the obtained results that the proposed hybrid model provides better prediction accuracy than the LS-SVM model.