Estudo online da dinâmica espaço-temporal de crimes através de dados da rede social Twitter

• Ano de defesa: 2015
• Autor(a) principal: Larissa Sayuri Futino Castro dos Santos
• Tipo de documento: Dissertação
• Tipo de acesso: Acesso aberto
• Idioma: por
• Instituição de defesa: Universidade Federal de Minas Gerais; UFMG
• Palavras-chave em Português: Árvore de decisão; Twitter; Mineração de texto; SVM; Naive-Bayes; Estatística; Estatística criminal; Mineração de dados (Computação); Probabilidades
• Link de acesso: http://hdl.handle.net/1843/BUBD-9VDJH7

To understand crime dynamics is essential for the development of public politics to control many types os crimes. In this study, we aim to map the crime occurrences at the state of São Paulo by collecting posts from the Twitter Social Web. Using the coollected data and machine learning techniques this study aim to classify, in an automatic way, the occurrences of crimes in the cited area. This way, we are able to dinamically visualize space time aspects of the crime distribution due to the possibility of real time collection of data. In this work, we present how to collect tweets and the machine learning methodology for the tweet classification. At first, we present and use three text classification techniques, known as, Naive-Bayes, Decision Trees and Support Vector Machines (SVM). Next, a cross validation study is performed for each technique and they are compared by classification eficiency and computational time.