小样本贝叶斯网络结构学习的KDE-CGA算法

KDE-CGA Algorithm of Structure Learning for Small Sample Data Bayesian Network

针对小样本数据条件下的贝叶斯网络结构学习,首先利用核密度估计(Kernel Density Estimation,KDE)对小规模样本数据进行拓展,然后引用云遗传算法(Cloud Theory-based Genetic Algotithm,CGA)对贝叶斯网络结构进行学习。通过优化改进核密度函数及其窗宽提高数据拓展效果;通过将云理论引入遗传算法中,自适应地改变交叉率和变异率,避免了算法局部寻优问题。仿真结果验证了该算法的有效性。

In view of learning the Bayesian network under the condition of the small sample data, this paper f irst ly made use of kernel density estimation to expand the small scale sample data, then adopted the cloud theory-based genetic al-go ti thm to learn the structure of Bayesian network. In order to improve the effect of data expanding, the paper discussed the way of improving the density funct ion and its window breadth. At the same t ime, the cloud theory was combined with genetic algo t ithm. We changed crosses rate and variation rate properly, avoided the problem of looking an excellent answer in a part. Simulation results show that the algorithm is effective and practical.