Prediction Method of Soil Oil Concentration Based on Wavelet Kurtosis

In order to solve the prediction problem of oil pollutant concentration on soil surface, oil spectral curve was obtained by laser induced fluorescence technology. Based on kurtosis, skewness and standard deviation, wavelet kurtosis was proposed as the quantitative parameter to predict the concentration of contaminated oil on soil surface. The random forest regression algorithm was used to compare and analyze three different oil products in the market. The experimental results show that the prediction accuracy of three kinds of oil concentration by random forest regression model is improved by 6.67%, 6.66% and 3.33% respectively. It provides some reference for the regression model of predicting the concentration of oil pollutants in soil surface.