A Gas Emission Prediction Model Based on Feature Selection and Improved Machine Learning

This paper proposed a gas emission prediction method based on feature selection and improved machine learning, as traditional models are neither accurate nor universally applicable. Through analysis, this identified 12 factors that affected emissions. A total of 30 groups typical data for outflow were standardized, after which full subset regression was used to categorize influencing into different regular patterns select 18 parameter sets. Meanwhile, nuclear principal component analysis (KPCA), an optimized model constructed where the dimensionality original reduced. An algorithm set hybrid kernel extreme learning (HKELM) least squares support vector (LSSVM). The performance parameters adopted in evaluated according certain metrics. By comparing results sets, final sequence could be obtained, composed optimal applied algorithm. showed HKELM outperformed LSSVM accuracy, running speed, stability. root meant square error (RMSE) 0.22865, determination coefficient (R2) 0.99395, mean absolute (MAE) 0.20306, percentage (MAPE) 1.0595%. Every index accuracy evaluation performed well had high-prediction wide application.