Prediction of uncertainties of gas emission quantity based on RVM

Abstract

In order to evaluate uncertainties of prediction results of gas emission, an estimation approach based on relevant vector machine was proposed. The sparse relevant support vector and its corresponding hyper parameters were calculated on sample space of gas emission by sparse Bayesian learning model. The mean and variance of prediction results were worked out, so probability distribution and confidence interval of prediction results of gas emission quantity can also be obtained. The analysis results show that the mean prediction error of three group testing samples is 1.74%, and real gas emission quantities are all in confidence interval of 97%. The prediction result is consistent with actual situation, it shows that the proposed approach can be used to get probability distribution of prediction result of gas emission, and has high prediction accuracy and requires less support vectors.

Keywords

• gas emission prediction
• uncertainty
• relevant vector machine
• probability distributio