Crime Prediction Using Support Vector Machine and Extracted Twitter Features
Jhonatan F. Sossa Rojo, Luís Alejandro Flétscher Bocanegra, Juan Felipe Botero, Natalia Gaviria Gómez · 2023
Citizen security in smart cities is a historical prob-lem that is addressed from several fronts. One of them is the implementation of intelligent solutions based on computational intelligence techniques that allow to develop preventive and/or reactive strategies to improve the life quality of citizens. This article presents a strategy that allows cities to predict criminal acts in specific spaces and times based on citizen participation through the Twitter social network. The proposed method consists of using a natural language processing model to extract features of the tweets made in the space and time of interest in order to predict if a criminal act will be committed in a later time window. The proposed method is able to predict future time windows where criminal acts will potentially be committed with a percentage of 87 % with a resulting accuracy of 77 %.