Research on an Identification Method for Gas Disaster Risk Based on the Selective Ensemble Classification Model.

Area of Science:

• Geosciences
• Computer Science
• Data Science

Background:

• Accurate gas disaster risk identification is crucial for mine safety.
• Existing classification models may have limitations in predicting complex disaster scenarios.

Purpose of the Study:

• To propose a novel selective ensemble classification model, CS-NDCF, for improving gas disaster risk identification accuracy.
• To evaluate the effectiveness of CS-NDCF compared to individual base classifiers.

Main Methods:

• Constructed nine base classifiers: backpropagation (BP) neural network, naive Bayes (NB), K-nearest neighbor (KNN), logistic regression (LR), decision tree (DT), support vector machine (SVM), SVM with cross-validation (SVMCV), random forest (RF), and gradient boosting DT (GBDT).
• Employed K-means clustering to group base classifiers by performance.
• Utilized a novel degree of combination fitness (CS-NDCF) to select optimal classifiers for the ensemble model.

Main Results:

• The CS-NDCF model demonstrated significant accuracy improvements over individual classifiers.
• Accuracy gains ranged from 1.20% (GBDT) to 34.83% (NB).
• CS-NDCF achieved superior forecasting results on actual mine monitoring data.

Conclusions:

• CS-NDCF is an effective method for gas disaster risk identification.
• The proposed model shows strong application value in real-world mining safety scenarios.