Abstract:Users increasing activity across various social networks made it the most widely used platform for exchanging and propagating information among individuals. To spread information within a network, a user initially shared information on a social network, and then other users in direct contact with him might have shared that information. Information expanded throughout the network by repeatedly following this process. A set of information that became popular and was repeatedly shared by different individuals was called viral events. Identifying and analyzing viral social events led to valuable insights into the dynamics of information dissemination within a network. However, more importantly, proactive approaches emerged. In other words, by observing the dissemination pattern of a piece of information in the early stages of expansion, it became possible to determine whether this cascade would become viral in the future. This research aimed to predict and detect viral events in social networks by observing granular information and using a deep survival analysis-based method. This model could play a significant role in identifying rumors, predicting the impact of information, and assisting in optimal decision-making in information management and marketing. Ultimately, the proposed method was tested on various real-world datasets from Twitter, Weibo, and Digg.

What problem does this paper attempt to address?

### What problem does this paper attempt to solve? This paper aims to solve the problem of how to predict and detect viral events in social networks. Specifically, the authors focus on predicting whether a piece of information will become a viral event (i.e., be widely shared by a large number of users) by observing the early stage of information dissemination. This involves the following key aspects: 1. **Identifying viral events**: Analyze the dissemination patterns of information in social networks to identify which pieces of information will become viral events. This is very important for understanding the dissemination dynamics of information in the network. 2. **Early - stage prediction**: More importantly, study how to predict whether it is possible for information to become a viral event in the early stage of information dissemination. This enables relevant parties to take measures in advance, such as dealing with rumors and optimizing marketing strategies. 3. **Handling censored data**: Since usually only part of the dissemination process can be observed, how to accurately predict the future dissemination situation from limited early - stage data is a challenge. The paper proposes a method based on deep survival analysis to solve this problem. 4. **Application areas**: - **Rumor detection**: Detect and manage rumors that may be widely spread as early as possible. - **Influence prediction**: Evaluate whether current news or content will become a hot topic, thereby influencing user behavior. - **Supply - demand analysis**: Predict the advertising effects of social network users to optimize the supply of goods and services. - **Marketing decision - making**: Help enterprises evaluate the effects of advertising or marketing activities. - **Promote the dissemination of beneficial information**: Identify high - impact news early to enhance social participation. 5. **Technical means**: Utilize deep learning and survival analysis models, especially by fitting survival functions to predict the time distribution of information dissemination, and based on this, perform classification to determine whether information will become a viral event. ### Main methods The paper proposes a method named VEDSA (Viral Event Detection using Deep Survival Analysis), which mainly includes two stages: - **γ - model (Survival Function Fitting)**: Use the LSTM network to fit the survival function and capture long - term dependencies and the correlation of time - series data. - **δ - model (Discriminator)**: Based on the fitted survival function, extract features through convolution and pooling layers, and finally perform binary classification through the fully - connected layer to determine whether information will become a viral event. ### Dataset and experimental results The paper conducted experiments on three real - world datasets, namely Twitter, Weibo, and Digg, to verify the effectiveness of the proposed method. The results show that, especially when using the Weibull distribution, the VEDSA model has excellent accuracy and F1 - score in multiple time periods. ### Summary This research, by combining deep learning and survival analysis, provides an effective method to predict viral events in social networks, especially suitable for early - stage prediction and handling censored data situations. This method not only helps to understand the dynamics of information dissemination but also provides strong support for various practical application scenarios.

Detecting Viral Social Events through Censored Observation with Deep Survival Analysis

Socialanalysis: A Real-Time Query And Mining System From Social Media Data Streams

COVID-19 Surveiller: toward a robust and effective pandemic surveillance system basedon social media mining

COVID-19 Surveiller: Toward a Robust and Effective Pandemic Surveillance System Based on Social Media Mining

Exploratory Analysis for Big Social Data Using Deep Network

A Real-Time Method to Predict Social Media Popularity

User Behavior Discovery in the COVID-19 Era through the Sentiment Analysis of User Tweet Texts

A Tutorial on Event Detection using Social Media Data Analysis: Applications, Challenges, and Open Problems

Ontology-based Data Collection for a Hybrid Outbreak Detection Method Using Social Media

COVID-19 and Big Data: Multi-faceted Analysis for Spatio-temporal Understanding of the Pandemic with Social Media Conversations

Event Early Embedding: Predicting Event Volume Dynamics at Early Stage

A Semi-automatic Method for Efficient Detection of Stories on Social Media

Early Identification of Pathogenic Social Media Accounts

Prediction and Characterization of High-Activity Events in Social Media Triggered by Real-World News

Identifying underlying influential factors in information diffusion process on social media platform: A hybrid approach of data mining and time series regression

#StayHome or #Marathon? Social Media Enhanced Pandemic Surveillance on Spatial-temporal Dynamic Graphs

Event Detection from Social Media for Epidemic Prediction

Information Contagion: an Empirical Study of the Spread of News on Digg and Twitter Social Networks

Predicting Viral Rumors and Vulnerable Users for Infodemic Surveillance

Early Warning Analysis for Social Diffusion Events

Predictive Analysis for Social Diffusion: The Role of Network Communities