Multi-label prediction method for lithology, lithofacies and fluid classes based on data augmentation by cascade forest

Predicting the lithology, lithofacies and reservoir fluid classes of igneous rocks holds significant value in domains CO2 storage evaluation. However, no precedent exists for research on multi-label identification rocks. This study proposes a data augmented cascade forest method prediction multilabel using 9 conventional logging features cores collected from eastern depression Liaohe Basin northeastern China. Data augmentation is performed an unbalanced training set synthetic minority over-sampling technique. Sample achieved by consisting predictive clustering trees. These structures possess adaptive feature selection layer growth mechanisms. Given necessity to focus all possible outcomes generalization ability method, simulated well model built then compared with 6 typical learning methods. The outperformance this evaluation metrics validates its superiority terms accuracy ability. consistency predicted results geological actual wells verifies reliability our method. Furthermore, show that it can be used as reliable means fluids. Document Type: Original article Cited as: Han, R., Wang, Z., Guo, Y., X., A, Zhong, G. Multi-label based forest. Advances Geo-Energy Research, 2023, 9(1): 25-37. https://doi.org/10.46690/ager.2023.07.04