An Initial Experiment on Associations between Crimes in Brazil

• Main Author: Gomes, Wesckley Faria
• Publication Date: 2020
• Other Authors: Colaço Júnior, Methanias, Prado, Kleber Henrique de Jesus
• Format: Article
• Language: por
• Source: Research, Society and Development
• Download full: https://rsdjournal.org/index.php/rsd/article/view/10078
• topic: Análise criminal; Criminalidade; Data science; Mineração de dados; Python e R juntos.; Análisis criminal; Crimen; Ciencia de los datos; Procesamiento de datos; Python y R juntos.; Criminal analysis; Crime; Data science; Data mining; Python and R together.

Context: Crime has been a problem around the world, causing damage to societies. Education, poverty, employment and climate are some of the factors that affect the crime rate, leading authorities to spend millions annually on actions to combat violence and strategic plans to prevent and reduce crime. Objective: Applying Data Science concepts to analyze government data related to crimes in Brazil. Method: Use of data mining techniques of association rules (AR), in a controlled experiment, to detect patterns between the types of crimes, as well as the relationship between the types of crime and the months of the year. Results: In the context of associations between crimes, the states with the most interesting rules were: Bahia, with 15 associations, São Paulo, with 12 associations, Goiás, with 11, and Paraná, with 9. Highlight for the association “Robbery Resulting in Death  Cargo Roberry”, found for the State of Bahia, which reached 99% confidence (0.99). In the scope of associations between crimes and months of the year, Paraíba had 2 associations, Maranhão, Rondônia and São Paulo, with 1 association each. Highlight for rule “December  Vehicle Steal”, found for the State of São Paulo, which reached a confidence of 84% (0.84). Conclusion: The results exposed in this research assist criminal analysts in the decision-making process to prevent and reduce crime in the country, since they can allow priority in inhibiting crimes that are antecedents of other occurrences within the same state or crimes that occur in the same period.