论文部分内容阅读
针对纯数据驱动的贝叶斯网络结构学习算法的准确度和效率较低的问题,提出了一种融合多信号流图模型与K2学习算法的贝叶斯网络自动建模方法。该方法利用多信号流图模型能够描述信号之间传递与依赖关系的能力,结合K2学习算法在结构学习中的优势,实现了专家知识与数据驱动方法有效融合的贝叶斯网络结构自动学习算法。通过与常用网络结构学习算法的对比实验证明,该融合算法显著降低了结构学习对学习范围和训练数据规模的要求,具有更高的学习准确度和运算效率。采用真实系统实例阐述了该融合算法的应用过程,验证了算法的实用性。
To solve the problem of low accuracy and low efficiency of pure data-driven Bayesian network structure learning algorithm, an automatic Bayesian network modeling method based on multi-signal flow graph model and K2 learning algorithm is proposed. This method uses the multi-signal flow graph model to describe the ability of signal transmission and dependence. Combined with the advantages of K2 learning algorithm in structure learning, an automatic Bayesian network structure learning algorithm that combines expert knowledge and data-driven method is realized . Compared with the commonly used network structure learning algorithms, this fusion algorithm has proved that the fusion algorithm can significantly reduce the learning scope and training data size requirements, and has higher learning accuracy and computing efficiency. The real system example is used to illustrate the application process of the fusion algorithm and verify the practicability of the algorithm.