Abstract: Social media is an essential data source for risk perception of complex social events. However, existing early warning methods mainly rely on popularity peaks, task-specific semantic labels, or observable propagation structures, often ignoring propagation rhythm changes prior to a crisis. This paper proposes an unsupervised online framework for social media crisis early warning. Aggregating heterogeneous behaviors into a unified intensity sequence, it detects anomalously accelerated propagation phases relative to historical baselines through causal smoothing, rate-of-change extraction, rolling quantile thresholds, persistence constraints, and cooling mechanisms. Compared to intensity-driven methods, it focuses on “whether the system is accelerating toward a risk state” rather than “whether it has reached a high popularity state;” compared to semantic- and structure-driven methods, it has weaker dependencies on annotations, ontologies, and complete propagation graphs. Strict causal replay experiments on the Russia–Ukraine crisis Weibo dataset and the Douban Movie Short Comments dataset demonstrate that the framework captures aggregative rising processes before key events early with low false alarm interference, exhibiting structural reusability across scenarios. Results indicate that detecting propagation dynamics changes offers a lightweight, interpretable, and transferable path for online early warning in complex systems.

社会雷达多普勒：基于社交媒体的危机早期预警多场景验证

[1]Alam F, Sajjad H, Imran M, et al., 2021. CrisisBench: benchmarking crisis-related social media datasets for humanitarian information processing. Proc 15^th Int AAAI Conf on Web and Social Media, p.923-932.

[2]Bian T, Xiao X, Xu TY, et al., 2020. Rumor detection on social media with bi-directional graph convolutional networks. Proc 34^th AAAI Conf on Artificial Intelligence, 34(1):549-556.

[3]Chung-I, 2020. Douban-Sentiment-Analysis. https://github.com/Chung-I/Douban-Sentiment-Analysis [Accessed on June 26, 2026].

[4]Diao QM, Jiang J, 2013. A unified model for topics, events and users on Twitter. Proc Conf on Empirical Methods in Natural Language Processing, p.1869-1879.

[5]Fung YR, Ji H, 2022. A Weibo dataset for the 2022 Russo–Ukrainian crisis.

[6]Gallotti R, Valle F, Castaldo N, et al., 2020. Assessing the risks of ‘infodemics’ in response to COVID-19 epidemics. Nat Hum Behav, 4(12):1285-1293.

[7]Gui JY, Ma ZN, Zhou H, et al., 2025. Deep anomaly detection of temporal heterogeneous data in AIOps: a survey. Front Inform Technol Electron Eng, 26(9):1551-1576.

[8]Han SJ, Zhang JS, Shen YQ, et al., 2025. FinSphere: a real-time stock analysis agent with instruction-tuned large language models and domain-specific tool integration. Front Inform Technol Electron Eng, 26(10):1822-1831.

[9]Imran M, Castillo C, Diaz F, et al., 2015. Processing social media messages in mass emergency: a survey. ACM Comput Surv, 47(4):67.

[10]Lasswell HD, 1948. The structure and function of communication in society. In: Bryson L (Ed.), The Communication of Ideas. Harper and Row, New York, p.37-51.

[11]Lau JH, Collier N, Baldwin T, 2012. On-line trend analysis with topic models: #Twitter trends detection topic model online. Proc COLING, p.1519-1534.

[12]Li X, Gao RJ, Wang R, 2025. Bursty event detection in social networks by integrating word features and Word2Vec model. Proc 4^th Int Conf on Intelligent Systems, Communications and Computer Networks, p.300-306.

[13]Liu JF, Wu YC, Liu ZX, 2025. A focused crawling strategy based on comprehensive priority evaluation of hyperlinks and improved Bayesian classifier. Front Inform Technol Electron Eng, 26(12):2569-2582.

[14]Liu YS, Chen Y, Zhang JY, et al., 2024. Toward an accurate mobility trajectory recovery using contrastive learning. Front Inform Technol Electron Eng, 25(11):1479-1496.

[15]Ma CB, Wang YX, Wu J, et al., 2025. Explicit and implicit data augmentation for social event detection. Proc 63^rd Annual Meeting of the Association for Computational Linguistics, p.8402-8415.

[16]Ma J, Gao W, Wong KF, 2017. Detect rumors in microblog posts using propagation structure via kernel learning. Proc 55^th Annual Meeting of the Association for Computational Linguistics, p.708-717.

[17]Mathieu J, Fulk M, Lorber M, et al., 2012. Social Radar Workflows, Dashboards, and Environments. RTO-MP-HFM-201, The MITRE Corporation.

[18]Parekh T, Mac A, Yu JR, et al., 2024. Event detection from social media for epidemic prediction. Proc Conf of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, p.5758-5783.

[19]Peng SC, Cao LH, Wang GJ, et al., 2025. A survey on textual emotion cause extraction in social networks. Dig Commun Netw, 11(2):524-536.

[20]Peng XY, Wu JR, Liu RM, et al., 2025. Rumor detection on social media with temporal propagation structure optimization. Proc 31^st Int Conf on Computational Linguistics, p.3865-3878.

[21]Qiu ZT, Ma CB, Wu J, et al., 2025. Text is all you need: LLM-enhanced incremental social event detection. Proc 63^rd Annual Meeting of the Association for Computational Linguistics, p.4666-4680.

[22]Ramakrishnan N, Butler P, Muthiah S, et al., 2014. ‘Beating the news’ with EMBERS: forecasting civil unrest using open source indicators. Proc 20^th ACM SIGKDD Int Conf on Knowledge Discovery and Data Mining, p.1799-1808.

[23]Schramm W, Roberts DF, 1954. How communication works. In: Schramm WL (Ed.), The Process and Effects of Mass Communication. University of Illinois Press, Urbana, p.3-26.

[24]Tang BH, Lin BD, Yan HL, et al., 2024. Leveraging generative large language models with visual instruction and demonstration retrieval for multimodal sarcasm detection. Proc Conf of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, p.1732-1742.

[25]Utmhikari, 2019. Douban Movie Short Comments Dataset. https://www.kaggle.com/datasets/utmhikari/doubanmovieshortcomments [Accessed on June 26, 2026].

[26]Wang FY, 2007. Toward a paradigm shift in social computing: the ACP approach. IEEE Intell Syst, 22(5):65-67.

[27]Weng JS, Lee BS, 2011. Event detection in Twitter. Proc 5^th Int AAAI Conf on Web and Social Media, p.401-408.

[28]Xie W, Zhu FD, Jiang J, 2016. TopicSketch: real-time bursty topic detection from Twitter. IEEE Trans Knowl Data Eng, 28(8):2216-2229.

[29]Xiong SF, Zhang GP, Fan XB, 2025. MAL: multilevel active learning with BERT for Chinese textual affective structure analysis. Front Inform Technol Electron Eng, 26(6):833-846.

[30]Zhang YZ, Wang MY, Wu YX, et al., 2025. DialogueLLM: context and emotion knowledge-tuned large language models for emotion recognition in conversations. Neur Netw, 192(1):107901.

[31]Zhou F, Xu X, Trajcevski G, et al., 2021. A survey of information cascade analysis: models, predictions, and recent advances. ACM Comput Surv, 54(2):27.

Similar articles