Quantitative analysis of coal industrial index based on data set partitioning method

Abstract

In order to further improve the accuracy of quantitative analysis of coal quality by laser induced breakdown spectroscopy (LIBS), the influence of data set partitioning method on quantitative model was studied. The spectral data of 40 different coal samples were collected, and the Support Vector Regression (SVR) model and random forest (RF) model were established by Random Selection (RS), Kennard-Stone (KS) and Sample Partitioning based on joint X-Y distances (SPXY), respectively. The prediction results of ash, volatile matter and calorific value under the two models were compared. The results show that the regression model established by SPXY method combined with RF algorithm has better fitting prediction performance. The predicted root mean square errors (RMSEP) of ash, volatile matter and calorific value are 1.8872, 1.4537 and 0.9020, respectively, and the mean relative errors (MRE) are 6.96%, 3.87% and 2.14%, respectively.