Use of neural networks to classify papers in systematic reviews

Abstract

The number of graduate students in Brazil is increasing every year. This growth is extremely necessary because research is fundamental to the development of a country and much of the world research is developed with the participation of graduate students. Typically, research begins with a literature review, and if the goal is to know the state of the art of a particular subject through a well-formulated and reproducible process, a systematic review can be used. However, reviews such as the systematic one tend to be quite rigorous, time-consuming and hard to be performed manually. The objective of this paper is the development of a method to automatically assist in the classification of papers to be included or excluded in a systematic review through an MLP neural network, maximizing the reading of papers that are of interest to the research. The proposed solution was evaluated with two datasets and the results were compared with those produced by two other classifiers. MLP had the best result among the methods tested in both datasets, corresponding to a good choice for this type of task.