AN ONLINE FRAMEWORK FOR CIVIL UNREST PREDICTION USING TWEET STREAM BASED ON TWEET WEIGHT AND EVENT DIFFUSION

Twitter is one of most popular Internet-based social networking platform to share feelings, views, and opinions. In recent years, many researchers have utilized the social dynamic property of posted messages or tweets to predict civil unrest in advance. However, existing frameworks fail to describe the low granularity level of tweets and how they work in offline mode. Moreover, most of them do not deal with cases where enough tweet information is not available. To overcome these limitations, this article proposes an online framework for analyzing tweet stream inpredicting future civil unrest events. The framework filters tweet stream and classifies tweets using linear Support Vector Machine (SVM) classifier. After that, the weight of the tweet is measured and distributed among extracted locations to update the overall weight in each location in a day in a fully online manner. The weight history is then used to predict the status of civil unrest in a location. The significant contributions of this article are (i) A new keyword dictionary with keyword score to quantify sentiment in extracting the low granularity level of knowledge (ii) A new diffusion model for extracting locations of interest and distributing the sentiment among the locations utilizing the concept of information diffusion and location graph to handle locations with insufficient information (iii) Estimating the probability of civil unrest and determining the stages of unrest in upcoming days. The performance of the proposed framework has been measured and compared with existing logistic regression based predictive framework. The results showed that the proposed framework outperformed the existing framework in terms of F1 score, accuracy, balanced accuracy, false acceptance rate, false rejection rate, and Matthews correlation coefficient.

Abstract licence: CC BY

33 more references, whose wording the paper’s licence keeps private.

A person counts once a day per paper. We keep no cookie and no address: a reader is a hash that changes every day. Any month under five reads shows no number.