Analysis of Lightgbm Model Accuracy and Validation Strategies for Predicting Oil and Gas Probability Maps
Main Article Content
Abstract
The accuracy of the LightGBM model in forecasting oil and gas probability maps had been evaluated based on MapOil analytic system. Integration of geological, geophysical, satellite and topographic map datasets in single spatial processing pipeline through MapOil streamline feature extraction, and probability assessment for each grid cell over exploration areas. The leaf-wise growth and advanced gradient boosting optimization criteria of LightGBM are capable of capturing the complex nonlinear relationships associated with subsurface properties. Differences among the various validation strategies are also examined to compare the robustness of predictions. Standard random cross-validation along with spatially informed cross-validation approaches, such as block and region-based validation, were utilized. Validation in space results in reduced information leakage and more realistic performance statistics, leading to a more informative generalisation of the model on untested areas. In summary, the application of LightGBM in combination with MapOil platform forms a powerful methodology for creating high-precision oil and gas probability maps which are able to improve the data driven decision making for hydrocarbon exploration.
Article Details
Issue
Section
How to Cite
References
[1] T. Zhang, H. Chai, H. Wang, T. Guo, L. Zhang, and W. Zhang, “Application of LightGBM for shear wave velocity estimation in carbonate layers,” Journal of Petroleum Science, 2021.
[2] W. J. Al Mudhafar, A. A. Hasan, M. A. Abbas, and D. A. Wood, “Predicting permeability in carbonate reservoirs using LightGBM and well logs,” Computers & Geosciences, 2020.
[3] A. Roustazadeh, B. Ghanbarian, M. B. Shadmand, V. Taslimitehrani, and L. W. Lake, “Machine learning approaches for estimating recovery factor in hydrocarbon reservoirs,” Journal of Petroleum Science and Engineering, 2019.
[4] G. Ke et al., “LightGBM: A highly efficient gradient boosting decision tree,” in Advances in Neural Information Processing Systems, 2017.
[5] J. Friedman, “Greedy function approximation: A gradient boosting machine,” Annals of Statistics, 2001.
[6] T. Chen and C. Guestrin, “XGBoost: A scalable tree boosting system,” in Proc. 22nd ACM SIGKDD, 2016.
[7] S. M. Lundberg and S.-I. Lee, “A unified approach to interpreting model predictions,” in Advances in Neural Information Processing Systems, 2017.
[8] T. Hengl et al., “SoilGrids1km — Global soil information based on automated mapping,” PLoS ONE, 2014.
[9] F. Chollet, Deep Learning with Python. Manning Publications, 2018.
[10] C. M. Bishop, Pattern Recognition and Machine Learning. Springer, 2006.
[11] T. Hengl and G. B. M. Heuvelink, Geostatistical Approaches for Soil Mapping. Springer, 2010.
[12] I. Goodfellow, Y. Bengio, and A. Courville, Deep Learning. MIT Press, 2016.
[13] T. Hastie, R. Tibshirani, and J. Friedman, The Elements of Statistical Learning. Springer, 2009.
[14] L. Breiman, “Random forests,” Machine Learning, 2001.
[15] C. E. Rasmussen and C. K. I. Williams, Gaussian Processes for Machine Learning. MIT Press, 2006.