基于贝叶斯网络的新媒体事件分类模型

为了对新媒体事件进行准确有效分类,提出一种将K-means算法和贝叶斯网络相结合的混合算法。该算法首先运用K-means算法将训练样本聚类,再根据聚类的结果,运用改进的贝叶斯网络对新媒体事件进行分类。其中,层次贝叶斯网络模型的构建,避免了贝叶斯网络在参数学习时陷入局部寻优;同时引入隐藏节点,更大程度满足了贝叶斯网络的条件独立假设。实验结果表明该算法效果明显。

Classification Model of New Media Events Based on Bayesian Network

To make an exact and effective classification of new media events, this paper puts forward a mixed algorithm combining K-means with Bayesian network. The algorithm firstly clusters training samples by K-means, and then classifies new media events through modified Bayesian network according to the clustering results. Hierarchy Bayesian network model built in this algorithm enables Bayesian network to get rid of local optimization in parameter learning, while the introduction of hidden nodes better meets the requirements of conditional independence assumption of Bayesian network. Experimental results show that the algorithm is effective.