Li Jinli, Jiang Ming, Qin Yunbai, Zhang Ran, Ling Sai Ho
College of Electronic Engineering, Guangxi Normal University, Guilin, China.
Business and Law School, Deakin University, Geelong, Australia.
Complex Intell Systems. 2023;9(1):115-131. doi: 10.1007/s40747-022-00729-2. Epub 2022 Jun 23.
The growth of population and the various intensive life pressures everyday deepen the competitions among people. Tens of millions of people each year suffer from depression and only a fraction receives adequate treatment. The development of social networks such as Facebook, Twitter, Weibo, and QQ provides more convenient communication and provides a new emotional vent window. People communicate with their friends, sharing their opinions, and shooting videos to reflect their feelings. It provides an opportunity to detect depression in social networks. Although depression detection using social networks has reflected the established connectivity across users, fewer researchers consider the data security and privacy-preserving schemes. Therefore, we advocate the federated learning technique as an efficient and scalable method, where it enables the handling of a massive number of edge devices in parallel. In this study, we conduct the depression analysis on the basis of an online microblog called Weibo. A novel algorithm termed as CNN Asynchronous Federated optimization (CAFed) is proposed based on federated learning to improve the communication cost and convergence rate. It is shown that our proposed method can effectively protect users' privacy under the premise of ensuring the accuracy of prediction. The proposed method converges faster than the Federated Averaging (FedAvg) for non-convex problems. Federated learning techniques can identify quality solutions of mental health problems among Weibo users.
人口增长以及日常各种高强度的生活压力加剧了人与人之间的竞争。每年数以千万计的人患有抑郁症，但只有一小部分人得到了充分治疗。诸如脸书、推特、微博和QQ等社交网络的发展提供了更便捷的交流方式，并提供了一个新的情绪宣泄窗口。人们与朋友交流，分享观点，拍摄视频来表达自己的感受。这为在社交网络中检测抑郁症提供了契机。尽管利用社交网络进行抑郁症检测反映了用户之间既有的联系，但较少有研究人员考虑数据安全和隐私保护方案。因此，我们提倡将联邦学习技术作为一种高效且可扩展的方法，它能够并行处理大量边缘设备。在本研究中，我们基于名为微博的在线微博客进行抑郁症分析。基于联邦学习提出了一种名为CNN异步联邦优化（CAFed）的新算法，以降低通信成本并提高收敛速度。结果表明，我们提出的方法在确保预测准确性的前提下能够有效保护用户隐私。对于非凸问题，所提出的方法比联邦平均（FedAvg）收敛得更快。联邦学习技术能够识别微博用户心理健康问题的优质解决方案。
