SONGKLANAKARIN JOURNAL OF SCIENCE & TECHNOLOGY

Volume 41, No. 06, Month NOVEMBER, Year 2019, Pages 1294 - 1300

Forecasting petroleum consumption using hybrid svr-de model emphasizing on optimal parameter selection technique

Thoranin Sujjaviriyasup

Abstract Download PDF

At present, liquid fuels remain the dominant source of transportation energy consumption all over the world. Accordingly, the future demand prediction of petroleum consumption is a very challenging task with regard to efficient supply management. In this paper, a hybrid SVR-DE model is developed and proposed to address the problem. The developed model takes ability of SVR model to formulate complex predictive function while DE algorithm is used to search the optimal parameters of SVR model. Moreover, the hybrid model is compared withboth ARIMA and SVR models as traditional single models. The experimental results indicated that the developed model outperforms traditional forecasting models based on MAE, MAPE, and sMAPE criteria. Furthermore, the forecast performance of hybrid model is significantly different from both traditional single models at 0.05 significance level. Consequently, the proposed model can be a promising tool for annual petroleum consumption.

Keywords

combined model, petroleum consumption, ARIMA, support vector regression, differential evolution

---