基于去噪自编码器的微博传播流行度演化研究

Research on the Popularity Evolution in Weibo Based on Denoising Auto-Encoder

从新闻的传播演化过程中析取出真实的传播效果,进而分析发现规律,是社交媒体网络传播研究的热点问题。但是现有数据中包含大量噪音数据,影响算法的泛化性和准确性,人工去噪过程耗时耗力。针对这个问题,本研究提出将去噪自编码器引入传播流行度分析,直接使用粗糙数据集挖掘共性传播特征,感知异常节点,削减处理和分析工作量。文中对流行度指标和网络模型进行了定义和设计,并利用微博的不同数据集,经过标准化和归一化后,对模型进行训练和检测。最终的实验结果证明了该模型的泛化性、可行性和可用性。

It is a hot issue in the study of the social media network diffusion to analyze the real effect from the news diffusion evolution process and then find out the law of discovery.However,the existing data contain a large number of noises,which affects the generalization and accuracy of the algorithm,and the manual denoising process is time-consuming and labor-consuming.To solve this problem,this study proposes to introduce the Denoising Auto-Encoder into the diffusion popularity analysis,which directly mines common diffusion features with a rough dataset to perceive abnormal nodes and reduce the workload of processing and analysis.In this paper,the popularity index is defined and the network model is designed,and the model is trained and validated by using different datasets of Weibo which has been standardized and normalized.The final experimental results prove the generalization,feasibility,and availability of the model.