College of Earth Sciences, Jilin University, Changchun 130061, China

In order to improve the precision and efficiency of logging lithology classification, based on the difference of lithology in log response and using random forest algorithm, the authors automatically classify and identify the lithology from the logging data. In addition, the importance of different variables from the lithology classification is analyzed. Taking the logging data of a uranium mining area at the southern end of Daqing placanticline in Songliao Basin as an example, the lithology of the logging data in the study area is identified. The prediction accuracy of random forest algorithm is 88.67%. Resistivity and spontaneous potential data show more importance in lithology classification. The results indicate that random forest algorithm is an efficient and accurate lithology classification method with higher prediction accuracy, compared with support vector machine method.