Rock burst is one of the common disasters in the field of engineering. The traditional rock burst prediction model assumes that the data distribution is fixed or stationary and that the training samples are independent and identically distributed. The model is prone to catastrophic forgetting problems during use. In this study, a Bayesian model with incremental learning (IL) properties based on the Averaged One-Dependent Estimator (AODE) is proposed. The Tangential stress (σθ), Uniaxial tensile strength (σt), Uniaxial compressive strength (σc), Stress coefficient (σθ/σc), Brittleness coefficient (σc/σt), and elastic energy index (Wet) are taken as the study parameters. A total of 382 sets of rock burst data were used as research parameters for training and testing, including the elastic energy index Wet; Evaluate the overall performance of the model using an 8-fold cross-validation method. Finally, compare the model with intelligent algorithms such as naïve Bayes (NB), k-nearest neighbor (KNN), Artificial Neutral Network (ANN), quadratic discriminant analysis (QDA), and single indicator prediction methods. The results show that the prediction accuracy of the model reaches 92.9%, and it has excellent stability, applicability, and generalization ability. Compared with the two data inputs, the accuracy of the incremental learning attribute model improved by 10.8%. Compared with non-incremental learning models, the prediction accuracy of incremental learning models has increased by 6.4% and 13.9%, respectively. Compared with other rock burst prediction models, this model has significantly better prediction performance than other models. The model was applied to a tunnel on the CZ railway for engineering verification, and the accuracy of the prediction results reached 100%, proving that the model can continuously learn while ensuring high prediction accuracy.

中文翻译：

---

一种基于平均独立估计量和增量学习的半朴素贝叶斯岩爆强度预测模型

岩爆是工程领域常见的灾害之一。传统的岩爆预测模型假设数据分布是固定的或平稳的，并且训练样本是独立同分布的。该模型在使用过程中很容易出现灾难性的遗忘问题。在这项研究中，提出了一种基于平均单相关估计量（AODE）的具有增量学习（IL）特性的贝叶斯模型。以切向应力（σθ）、单轴拉伸强度（σt）、单轴压缩强度（σc）、应力系数（σθ/σc）、脆性系数（σc/σt）和弹性能量指数（Wet）作为研究参数。共使用382组岩爆数据作为训练和测试的研究参数，包括弹性能量指数Wet；使用8折交叉验证方法评估模型的整体性能。最后，将该模型与朴素贝叶斯（NB）、k近邻（KNN）、人工神经网络（ANN）、二次判别分析（QDA）和单指标预测方法等智能算法进行比较。结果表明，该模型的预测精度达到92.9%，具有良好的稳定性、适用性和泛化能力。与两次数据输入相比，增量学习属性模型的准确率提高了10.8%。与非增量学习模型相比，增量学习模型的预测精度分别提高了6.4%和13.9%。与其他岩爆预测模型相比，该模型的预测性能明显优于其他模型。该模型应用于长广铁路某隧道进行工程验证，预测结果准确率达到100%，证明该模型能够持续学习，同时保证较高的预测精度。