Analysis of social networks and filtering of Arabic crime tweets based on an intelligent dictionary using a genetic algorithm

Abstract

Preserving a robust online community poses a significant difficulty due to the unrestricted flexibility members have in expressing themselves and behaving. This issue can be remedied through the implementation of user behavior monitoring and analysis, followed by the implementation of suitable actions. The objective of this research is to develop an intelligent dictionary using the genetic algorithm to identify and categorize Twitter posts related to criminal activities by collecting data from tweets. Once the data is preserved, graph analysis techniques are employed to evaluate interactions between users. Next, the user behavior is analyzed using metadata analysis, whereby the chronology associated with each user profile is obtained. Furthermore, the study analyzes the behavioral patterns of users over time. Afterward, a method based on rules is used to create a structure for aspect-based analysis of sentiment. This method assesses the subjectivity of the input text, distinguishing between factual information and personal opinions. Additionally, transformer-based sentiment analysis determines whether the tweet evokes positive or negative sentiment. Furthermore, the task involves constructing a model that can accurately categorize a tweet based on its relevance to criminal activity. Ultimately, the intelligent dictionary is utilized to identify and isolate anomalous behavior by selecting provocative profiles. A Twitter profile exhibiting significant similarity with criminal cases presents a perilous risk to society.