A Novel Hybrid Technique of Integrating Gradient-Boosted Machine and Clustering Algorithms for Lithology Classification

The significant body of research on lithology identification in recent years has laid emphasis on the improvement of classification performance using hybrid machine learning methods. To the best of our knowledge, a hybrid lithology classification model that integrates clustering results of well log data has not been developed. This study, therefore, exploits the advantage of incorporating results from clustering well log data into 2 and 3 groups using K-means and Gaussian mixture models (GMM) to construct a more accurate gradient-boosted machine (GBM) lithology model. The findings of the study showed that improved performance in terms of classification accuracy rate was achieved by the K-means-based GBM classifiers. In addition, GMM-based GBM established an enhanced performance when the developed classifiers were tested on the entire dataset. A rigorous examination of the confusion matrices generated by the classifiers further revealed that the increase in the performance from the clustering-based hybrid GBM models was attributed to the improvement in recognizing mudstone and siltstone, which represents the main lithofacies that are found in the South Yellow Sea's southern Basin. The findings from the present paper demonstrate that a clustering-based hybrid GBM model can handle new independent lithofacies classification better than GBM.