Detecção de spam em mensagens SMS utilizando aprendizagem de máquina

Abstract

SMS (Short Message Service) is still one of the simplest and most practical mobile communication services to reach consumers, regardless of the connection to an Internet network or the capacity of the handsets. Some applications provide resources for sending SMS messages, but because they are present on the Internet, there is room for malicious users to use them to send spam. In the field of Artificial Intelligence, areas such as Machine Learning and language studies may prove to be great allies in the development of systems that aid in the filtering of spam messages. In this study, it is shown how improvements in the performance of the Bayesian Naive classification algorithm were reached, using it to classify SMS messages supported by Artificial Neural Networks and Word Embedding vectors, used to predict and generalize probabilities for words that were not used in the training of the classifier.