Abstract
Abstract A fracture network model is a powerful tool for characterizing fractured rock systems. In this paper, we present the fracture network model by integrating a machine learning algorithm in two-dimensional setting to predict the natural fracture topology in porous media. We also use a machine learning algorithm to predict the fracture azimuth angle for the natural fault data from Kazakhstan. The results indicate that the fracture network model with LightGBM performs better in designing a fracture network parameter for hidden areas based on data from the known area. In addition, the numerical result of the machine learning algorithm shows a good result for randomly selected data of the fracture azimuth.