改进的组合优化决策树谣言判别方法研究

Study on Rumor Discrimination Based on Improved Combinatorial Optimization Decision Tree

随着互联网承载的信息量逐步增长,如何对网络中的谣言做出准确的识别变得越来越重要,现有谣言判别方法都建立在实验样本和训练样本的数据结构相同的前提下,而现实中谣言数据的属性结构并不完全一致会导致谣言判别模型的准确率受数据属性结构的影响较大。针对这个问题,对组合优化决策树算法(CODT)的建树算法做出了相应改进并提出一种E-CODT算法,实验结果表明,在实验样本与训练样本集数据结构不同时,改进后的算法表现出对谣言的判别更高的准确率、普适性以及突出的现实应用价值。

The amount of information carried by Internet is increasing gradually, how to make accurate identifica- tion of rumors in the network has becoming more and more important. The existing methods of rumor identification are based on the same data structure as the experimental samples and the training samples. In reality, the attribute struc- ture of the rumor data is not consistent, which leads to that the accuracy of the rumor discrimination model is greatly affected by the structure of data attributes. This article aims at this problem, the algorithm of combinatorial optimiza- tion decision tree algorithm （CODT） is improved, and an E - CODT algorithm is proposed. The experimental results show that the improved algorithm has higher accuracy, universality and prominent practical application value when the data structure of experimental samples and training samples is different.