Identifying On-site Users for Social Events: Mobility, Content, and Social Relationship

Aim:

to identify on-site users of an event, from whom we could gather valuable information regarding the process of events and investigate suspects when an event is associated with crime or terrorist.

Existing System:

To identify on-site users, the traditional and straight forward way is to search users whose tweets are geo-tagged and located within a certain range (e.g., 200 meters) of the center location of a target social event.

Drawback:

·         only 34% of Twitter users have meaningful location information in their profiles for on-site user identification and less than 1% of Twitter users tag their tweets with GPS location [2, 3].

·         The sheer volume of social media users and their posts and the lack of user location information make it particularly challenging to identify on-site users.

Proposed System:

Fused Feature Gaussian Process Regression (FEGOR) model, which exploits three influential factors in social networks for on-site user identification: mobility influence, content similarity, and social relationship.

Objectives:

·         propose a Fused fEature Gaussian prOcess Regression (FEGOR) model that combines Mobility Influence, Content Similarity, and Social Relationship to estimate the relative location of users w.r.t events, named as User-Social Event Distance, which is used to identify on-site users.

·         Method transforms all absolute GPS locations into relative distances between users and social events. Based on such location projection and transformation, not only accomplish the goal of identifying on-site users for social events, but also protect the individual location privacy in a coarse-grained level.

Methodology:

System Architecture:

the first step of the framework is data collection and preprocessing, which consists of two main functions: Social Event Localization and Event Related User Collection using hot words.

Social Event Modeling:

 identify the event participants as a subset of active users who have tweeted about the event and their locations are within certain range (e.g., 200 meters) of the event center. Using this list of known participants, we can extract features of the event from their historical locations, tweets, and user profiles, resulting in Collective Mobility Patterns, Social Event Topic, and Attendee Information respectively.

Event-related User Modeling

Event-related users are those who have tweeted about the target social event, which indicates that these users are interested in, and potentially participate in, the specific social event. For each event-related user whose location is unknown during the event time window (i.e., normal users), we would like to estimate the distance between these users and the social event.

User-Social Event Distance Estimation:

propose a Fused fEeature Gaussian prOcess Regression (FEGOR) model to estimate the User-Social Event Distance. And an information entropy based genetic algorithm is proposed for parameter learning.

Using the three distinct features extracted from both the target social event and its event-related users, we calculate the three factors that may contribute to estimating the user social event distance: mobility influence (FMI ), content similarity (FCS) and social relationship (FSR).We then combine these three factors using the proposed FEGOR model to estimate User-Social Event Distance, along with an information entropy based genetic algorithm for parameter learning

to identify the weights of different features.

instead of determining the absolute location of each user, aim to estimate the relative location, i.e., the distance between the user and the event, which refer to as the User-Social Event Distance.

three features that can be helpful to estimate the User-Social Event Distance:

 1) Mobility Influence. This is one key factor that determines whether a user will attend or be interested in a target social event. This feature measures the historical trajectory similarity between a specific user and all participants of an event. Intuitively, if a user has similar trajectories as event participants, it could be inferred that the user would attend or be close to the same event.

 2) Content Similarity. This feature quantifies the textual similarity between a user’s tweets and event topic. Each social event can usually be associated with a specific theme that describes the event topic. If a user is near the event, he/she may tweet more about the event, leading to a higher content similarity than those who are far away from the event.

 3) Social Relationship. A user’s social relationship also plays an important role when it comes to attending a social event. If a user’s friends are among the ones attending the event, then it is more likely that the user would attend the same social event, compared with other users whose friends are not on the list of event participants.

Future Enhancement:

plan to expand our model beyond Twitter and consider other types of social network platforms.

Conclusion:

transform all absolute location data into distance values, not only achieve goal, but also protect users’ location privacy in a coarse-grained level. Offers new insights into the motivating factors of users attending social events that contribute to estimating the user-social event distance.
For additional details comment below with requirements.