Intelligent prediction of development height of water-conducting fracture zone based on the SSA-RF model

In the coal mining process, the development height of the water-conducting fracture zone directly affects the water hazard risk of the mine, and accurately predicting the development height of the water-conducting fracture zone is crucial for ensuring mining safety and efficient resource extraction. However, the geological environment under thick loose layers and thin bedrock is complex, leading to significant limitations in the application of traditional prediction methods such as empirical formulas and theoretical analysis. In order to accurately predict the development height of the water-conducting fracture zone under thick loose layers and thin bedrock, a composite prediction model (SSA-RF) based on Sparrow Search Algorithm (SSA) and Random Forest (RF) was developed. The model iteratively optimizes the hyperparameters of the Random Forest regression model using the Out-of-Bag (OOB error), quickly determining the optimal hyperparameters for the model. The trained optimal model was then evaluated using a 10-fold cross-validation experiment. The results showed that the model achieved an R² value of 0.941, Mean Squared Error (MSE) of 31.241, and Mean Absolute Error (MAE) of 3.56, outperforming other prediction models such as SVM, BP-Network, Lasso, Elastic-Net, and Ridge in all three key performance indicators. The relative error modulus was extremely small, and the interquartile range was notably narrow. The SSA-RF model demonstrated high stability and consistency in repeated experiments. Furthermore, the SSA-RF model conducted an importance analysis of the 12 influencing factors. Through importance ranking charts and correlation heatmaps, the impact of key factors such as mining height, mining method, water pressure of the bottom aquifer, and thickness of the bottom clay layer on the development height of the water-conducting fracture zone was revealed. This validated that the SSA-RF model can enhance the reasonableness of the RF model's importance ranking of predictive indicators and provides a strong theoretical foundation for accurately predicting the development height of the water-conducting fracture zone under thick loose layers and thin bedrock conditions.