Comparative analysis of machine learning models for the detection of messages considered as online grooming

Abstract

This article presents a comparative study between machine learning algorithms to perform the automatic classification of text considered as grooming, for which different libraries of the Python programming language are used that allow the analysis and treatment of the data set, for the subsequent training of a predictive model with these same data divided into a training group and a validation group to check the accuracy of the classification model. The main objective of this study is to determine which algorithm has a better performance for the classification of this type of text, comparing the efficiency of the algorithms with the data set through different independent evaluation metrics. As a result of the comparison between the text classification algorithms, it is concluded that the Support Vector Machines model shows better performance in terms of the metrics obtained, obtaining an accuracy of 90%, and also offers greater accuracy for text classification. even with a high dimension of training data and without consuming many memory resources.