主流价值信息传播模型仿真数据集

本数据集在Windows、Mac OS X、Linux操作系统下均可读取，可根据需求进一步设计信息传播模型来研究主流价值信息的传播规律。数据集中标题是新闻内容的高度浓缩，是用户点击及后续发生评论行为的重要文本特征。评论行为是新闻是否流行的重要体现。时间序列能够体现用户评论行为在时间维度上的演变规律，体现了新闻流行度的随时间的变化。利用新闻标题文本与时间序列信息结合深度神经网络技术可以实现主流价值信息的流行度预测任务。在实际应用中，将新闻标题输入至一维CNN获取标题的语义特征，将时间序列输入至循环神经网络获取新闻的序列信息特征，将获取的时间序列特征和语义特征注入到注意力单元中动态学习两者的权重，最后通过全连接层输出流行度预测的结果。

This dataset is readable across Windows, Mac OS X, and Linux operating systems, and can be used to further design information propagation models for studying the dissemination rules of mainstream value-oriented information. In the dataset, news headlines, which are highly condensed summaries of news content, serve as critical textual features influencing user clicks and subsequent comment behaviors. Comment behaviors are important indicators of whether a news article goes viral. Time series data reflect the temporal evolution pattern of user comment behaviors, as well as the dynamic changes in news popularity over time. Combining news headline texts and time series information with deep neural network technologies enables the popularity prediction task for mainstream value-oriented information. In practical applications, news headlines are fed into a 1D CNN to extract semantic features, while time series data are input into a Recurrent Neural Network (RNN) to capture sequential information features of the news. The extracted time series features and semantic features are then fed into an attention unit to dynamically learn their respective weights, and finally, the popularity prediction results are output through a fully connected layer.

搜集汇总

背景与挑战

背景概述

该数据集是一个用于研究主流价值信息传播规律的仿真数据集，包含新闻标题和时间序列数据，适用于多种操作系统。通过结合一维CNN和循环神经网络技术，可以实现新闻流行度的预测任务。