GA-RF-based lithologic identification of volcanic rocks in electrical imaging logging
Aiming at the problem that it is difficult to accurately identify the lithology of complex volcanic rocks using conventional logging data,this paper proposes a GA-RF(genetic algorithm-random forest)based method for volcanic rock lithology identification using electric imaging logging.Firstly,the four texture features of energy,contrast,correlation and homogeneity of the electric imaging logging image are extracted by grey level co-occurrence matrix(GLCM)method,and the three texture features of roughness,contrast and orientation of the image are extracted by the Tamura method,and the texture feature dataset is established;then,the feature dataset is subjected to feature fusion,dimensionality reduction,and the feature samples are balanced by the ADASYN over-sampling algorithm,which reduces the impact of sample imbalance on the algorithm.imbalance on the algorithm;finally,the parameters of Random Forest algorithm are optimized by genetic algorithm to construction of volcanic rock Lithology identification model based on GA-RF(hereinafter referred to as GA-RF model)and compare it with the three algorithms of Random Forest,GBDT and LightGBM.The results of instance analysis show that the accuracy of GA-RF model can reach about 94％,which is much higher than the three comparison algorithms.The method effectively improves the accuracy and speed of volcanic rock lithology recognition,which can provide a reference for the sample imbalance problem as well as the lithology recognition by logging methods.
国家科技期刊平台
NSTL
万方数据
