Criminal Processing Investigation Support System (CPISS) Sistema de Suporte a Decisões em Investigações Criminais
Carregando...
Citações na Scopus
Tipo de produção
Trabalho de Conclusão de Curso
Data
2024-12-11
Autores
Bonnet, Jean-Luc
Carrasco, Lucas da Silva
Santos, Victor Alexandre dos
Carrasco, Lucas da Silva
Santos, Victor Alexandre dos
Orientador
Rodrigues, Paulo Sérgio Silva
Periódico
Título da Revista
ISSN da Revista
Título de Volume
Citação
Texto completo (DOI)
Palavras-chave
criminal,graph,Bayesian models
Resumo
Com o crescente número de casos criminalísticos e suas complexidades, os agentes investigativos enfrentam muitos desafios que podem diminuir a eficácia nas soluções desses casos, deixando até mesmo muitos casos insolúveis. Atualmente, existem diversas tecnologias para auxiliar esses investigadores, bem como ampliar a base de dados de investigação e até mesmo resolver casos, o processo investigativo depende de diversas fontes de dados, como, por exemplo: testemunhas, vítimas, redes sociais, câmeras de vigilância, entre outros, tornando a análise dessas informações uma tarefa cada vez mais desafiadora. Assim, este projeto indica os suspeitos com a maior probabilidade de ser o culpado e indica uma fila de suspeitos, auxiliando os agentes investigativos a encontrar um caminho mais assertivo em suas investigações. Este modelo utiliza correlações entre os indivíduos envolvidos no caso e um conjunto de fatos que serão quantificados para gerar sua importância, definindo assim um ranqueamento final das pessoas envolvidas. Portanto, o objetivo deste projeto é, a partir de fatos coletados por agentes investigativos de um crime, correlacionar esses fatos, ampliar novos fatos, criar novas correlações e encontrar sugestões de caminhos que levem a investigações mais assertivas, esses caminhos, por sua vez, são as filas de suspeitos entre a pessoa com maior probalidade de ser culpado e a vítima. A metodologia proposta foi testada em diversos casos reais, que foram adaptados com o objetivo de controlar os parâmetros para análise, obtendo resultados promissores em casos complexos e bem conhecidos como o da “Deputada Flordelis”, que obteve uma precisão equivalente a 79,22%; casos com baixa complexidade, como do chamado “Maníaco do Parque”, que obteve 100% de precisão; e de complexidade média, como do chamado “Fera da Penha”, que obteve cerca de 59,21% de precisão, mostrando que o método sugerido é eficiente, e pode auxiliar os investigadores em suas investigações criminais.
With the increasing number of criminal cases and their complexities, investigative agents face many challenges that can reduce the effectiveness in solving these cases, leaving even many cases unsolved. Currently, there are various technologies available to assist these investigators, as well as to expand the investigation database and even solve cases. The investigative process depends on various data sources, such as witnesses, victims, social media, surveillance cameras, among others, making the analysis of this information an increasingly challenging task. Thus, this project suggests suspects with the highest probability of being guilty and provides a list of suspects, helping investigative agents to find a more assertive path in their investigations. This model uses correlations between individuals involved in the case and a set of facts that will be quantified to determine their importance, thus defining a final ranking of the people involved. Therefore, the goal of this project is, based on facts collected by investigative agents of a crime, to correlate these facts, expand new facts, create new correlations, and provide suggestions for paths that lead to more assertive investigations. These paths, in turn, are the lists of suspects, ranging from the person with the highest probability of being guilty to the victim. The proposed methodology was tested on several real cases, which were adapted to control the parameters for analysis, obtaining promising results in complex and well-known cases, such as the “Deputada Flordelis” case, which achieved an accuracy of 79.22%; cases with low complexity, such as the “Maníaco do Parque” case, which achieved 100% accuracy; and medium complexity cases, such as the “Fera da Penha” case, which achieved an accuracy of around 59.21%, showing that the suggested method is efficient and can assist investigators in their criminal investigations.
With the increasing number of criminal cases and their complexities, investigative agents face many challenges that can reduce the effectiveness in solving these cases, leaving even many cases unsolved. Currently, there are various technologies available to assist these investigators, as well as to expand the investigation database and even solve cases. The investigative process depends on various data sources, such as witnesses, victims, social media, surveillance cameras, among others, making the analysis of this information an increasingly challenging task. Thus, this project suggests suspects with the highest probability of being guilty and provides a list of suspects, helping investigative agents to find a more assertive path in their investigations. This model uses correlations between individuals involved in the case and a set of facts that will be quantified to determine their importance, thus defining a final ranking of the people involved. Therefore, the goal of this project is, based on facts collected by investigative agents of a crime, to correlate these facts, expand new facts, create new correlations, and provide suggestions for paths that lead to more assertive investigations. These paths, in turn, are the lists of suspects, ranging from the person with the highest probability of being guilty to the victim. The proposed methodology was tested on several real cases, which were adapted to control the parameters for analysis, obtaining promising results in complex and well-known cases, such as the “Deputada Flordelis” case, which achieved an accuracy of 79.22%; cases with low complexity, such as the “Maníaco do Parque” case, which achieved 100% accuracy; and medium complexity cases, such as the “Fera da Penha” case, which achieved an accuracy of around 59.21%, showing that the suggested method is efficient and can assist investigators in their criminal investigations.