Abstract
In this study, the Xgboost method is employed for TOC estimation in mixed carbonate and siliciclastic shale from the Hashan area, Junggar Basin. The results show that this approach is effective for TOC estimation in this area although the model performance is not very excellent with a correlation coefficient of 0.54 between measured TOC and predicted TOC values, likely due to a small samples dataset. Therefore, the PCA method is applied to debase dimension of well log data from five dimensional to two-dimensional data, which enhances the correlation coefficient between the predicted and measured TOC from 0.54 to 0.68. Based on the model, the isopleth maps of TOC distributions in Fengcheng Formation were redrawn showing two shale oil exploration targets, which likely correspond to two depositional centers of this strata. All the same, the model in this work provides reliable data for shale oil evaluation in the study area and a good example under similar geological setting.