| 摘 要: 社交媒体的普及促进了信息传播,也不可避免地为谣言泛滥提供了传播途径。用户是谣言传播的关键一环,用户间的差异性使他们对信息真实性做出不同判断,但差异性带来的影响很少在检测环节被考虑。因此,考虑差异个体在信息判别上的表现不同,将其纳入识别指标。针对以往数据集陈旧、识别性能不佳的问题,采集近两年的真实数据,利用贝叶斯优化后的LightGBM算法进行识别。结果表明用户差异可以有效识别谣言,且贝叶斯改进的LightGBM算法比其他算法表现出了更好的性能。 |
| 关键词: 社交媒体 谣言检测 个体差异 LightGBM 贝叶斯优化 |
|
中图分类号: TP391
文献标识码: A
|
| 基金项目: 国家社会科学基金资助项目(22BGL240,19BJY099) |
|
| Research on Social Media Rumor Detection Based on Ensemble Learning |
|
WU Xian,NI Jing
|
(Business School, University of Shanghai for Science and Technology, Shanghai 200093, China)
wydhcwyjs@163.com; nijing501@126.com
|
| Abstract: The popularity of social media has facilitated information dissemination, and inevitably provided a channel for the proliferation of rumors. Users serve as critical agents in rumor propagation, and their inherent
differences lead to varied judgments regarding information authenticity. However, such differences are rarely incorporated into detection frameworks. This study addresses this gap by integrating the distinct performance of heterogeneous individuals in information discernment as recognition metrics. To overcome limitations of outdated datasets and suboptimal identification performance in prior research, we collected real-world data from the past two years. Utilizing the LightGBM algorithm optimized with Bayesian methods for detection, our results demonstrate that user differences effectively identify rumors. Furthermore, the Bayesian-optimized LightGBM outperforms other algorithms in detection performance. |
| Keywords: social media rumor detection individual differences LightGBM Bayesian optimization |
