LOCAL GOVERNMENT UNIT COMMON FELONY CLASSIFICATION: A TEXT MINING AND NATURAL LANGUAGE PROCESSING APPROACH

Abstract

The main objective of this research paper was to develop a model for crime classification in local government units like barangay in the Philippines. The study included natural language processing. Text mining process was applied in this research to test and evaluate the accuracy of three (3) different machine learning algorithms which are all famous for text classification such as naïve bayes, support vector machine (SVM) and neural networks. The blotter or complaint records of barangay Mayondon in Los Baños, Laguna, Philippines was used as a dataset for model development. Upon developing the model, it is found out that among the three algorithms, the neural network got the highest accuracy in terms of classifying crime in a blotter or crime report. It is also found in this paper that the males have the most probability to commit crime than the females. Hence, it is shown that physical injury is the most common crime committed in a barangay.