析取置信规则库系统参数优化的深度神经网络模型

基于深度学习的置信规则库系统（BRB-DL）比传统的置信规则库系统（BRB）具有更好的推理精度和适用性。然而，现有的BRB-DL在参数优化方面存在可移植性不足、应用效率低等局限性。鉴于此，本文将深度神经网络与析取置信规则库结合，有效减少了模型的规则和参数的数量，并引入梯度下降算法优化模型参数，提高了模型构建和优化的效率。最后，通过非线性函数的拟合，北京市空气质量污染预测和多个UCI公共分类数据集的实验对本文提出的方法进行验证，并将实验结果与现有的置信规则库系统和传统的机器学习方法进行了对比。结果表明，本文提出的方法比传统的方法具有更高的推理精度和更快的训练速度。

Deep neural network model for parameter optimization of disjunctive belief rule base system

Belief rule base based on deep learning (BRB-DL) has better inference accuracy and applicability than the belief rule base (BRB). However, the existing parameter optimization methods have limitations such as insufficient portability and inefficiency in the application. Therefore, for the above problems, this paper combines deep neural network with disjunctive belief rule base, which reduces the number of rules and parameters and improves the efficiency of model construction and optimization via the gradient descent algorithm. Finally, this paper fits nonlinear functions, Beijing air pollution prediction and experiments on public datasets from UCI, and compares the experimental results with some existing EBRB systems and conventional machine learning methods. The results show that the proposed method in this paper has higher accuracy and faster training velocity than the conventional methods.