Automatic Categorization of Answers by Applying Supervised Classification Algorithms to the Analysis of Student Responses to a Series of Multiple Choice Questions

Abstract

In recent years there has been a growing interest in machine learning for the classification and categorization of documents, texts and questions. This allows automating processes that if done with the intervention of the human being could have a high cost in time, and opens the doors for its implementation with inclusive systems for students with physical disabilities. This article describes a research work that uses data mining techniques to obtain classifiers that automatically identify the correct answers expressed by students, and these answers are then associated with a question with different options that are part of the process of evaluating the knowledge acquired by students during their formative process. In view of these consideration, where each question had multiple feasible, where each question had multiple feasible options to be selected; however, each question had only one correct answer. The answers are given by the students of the Open and Distance Modality of the Universidad Tecnica Particular de Loja were transcribed, with a total of 12960 transcriptions of the verbal answers obtained from the students. The results obtained by means of different classification algorithms were presented, analyzed and compared; giving as a result that the neural networks and the support vector machine (SVM) were the best to classify with an average percentage of 97% of success.