Modeling and applying an analytical process based on social web exploration to support decisions in public security

Abstract

Public security is a critical sector of Public Administration, with direct repercussions on the functioning and well-being of society; it is a potential user of tools associated with Data Science and Artificial Intelligence to assist decision-making and problem-solving. Its activities are characterized by a complex network of operations involving cycles of planning, monitoring, and carrying out actions aimed at preventing or correcting problems that may affect the security of people and public property. The social web makes information available through digital social networks, personal sites, news sites, blogs, emails, and digital forums, which can be mined for information extraction. In the context of public security, it is possible to carry out a series of processes such as the identification of criminal messages, the detection of events or social movements with the potential to cause damage to public property or people, retrieval of information for use in investigative processes, supporting forensic actions or judicial decisions, and the detection of people's opinions, feelings, and emotions about the actions taken by the agencies that promote security. This thesis aims to propose and apply an analytical process involving textual data sources from the social web to support activities and decisions within the scope of public security management, defining a framework for the entire process of extraction, storage, treatment, analysis, and visualization of the large volumes of information that can be extracted from these sources. To this end, several tools are used in the following sequence: (i) web scraping to obtain texts associated with public security issues; (ii) storage of texts in specific formats and appropriate bases, creating corpora (text sets); (iii) treatment or pre-processing of texts using natural language processing, to eliminate unwanted noise that could impair analysis; (iv) data analysis with Artificial Intelligence tools, specifically Machine Learning and its branches, to detect patterns, as in the case of the analysis of feelings or opinions; (v) visual presentation, through friendly graphics, enabling managers/decision makers to have an adequate understanding of the phenomenon under analysis. Therefore, the potential impacts of the research concern the generation and application of instruments for the rescue, structuring, and analysis of information extracted from the social web on topics of interest related to public security. With the analytical framework, it becomes possible to demonstrate, for example, the evolution of posts on a topic and where they were generated, helping to identify who generated them and other people mentioned, enabling the application of the results in the strategic decisions of the public security management, resulting in actions to improve the services offered to the population and in the fight against disinformation that may be associated.