Implication of Bayesian Technique to Improve Econometric Model Forecasts

This study presents the implication of Bayesian technique on simple statistical and econometric forecasting models to improve the forecast performances of the models. We consider a nonlinear, non-stationary time series of household electricity demand to demonstrate the Bayesian implication to statistical techniques. In this forecasting process, the electricity demand is considered a function of electricity price, household use of electric appliances, personal income, number of households, and urban conditions. We applied the Bayesian statistical technique on a multivariate linear regression model to predict the parametric values of the regression demand model. In the Bayesian process, the forecast is generated from the inferences of marginal posterior distribution of the model parameters obtained by Markov Chain Monte Carlo simulation process. Forecast result is tested and compared with actual data and two alternate models. The Bayesian model is proven to be an effective forecasting method with a great flexibility and capacity to solve multi-dimensional time series models by continuously updating the estimated parameter values as the demand changes over time. Test results indicate that Bayesian implication has significantly improved the performance of regression model. In the comparison, the error index reveal that Bayesian model outperform the forecast obtained by an artificial neural network model and a classical regression model. Application of Bayesian approach has proven to be efficient in predicting the trend changes and future proclivity of the large-scale multivariate time series data.