Imbalanced Learning Techniques for Improving the Performance of Statistical Models in Automated Essay Scoring
Aluízio Haendchen Filho, Fernando Concatto, Jonathan Nau, Hércules Antônio do Prado, Diego Oscar Imhof, Edílson Ferneda · Procedia Computer Science · 2019
The problem of class imbalance occurs in many application domains, as in the case of essays. The imbalance of the number or samples among the classes presents a problem for predictive algorithms, both for classification or for regression. In this paper, we present imbalanced learning techniques for improving the performance of statistical models in automated essay grading. We analyze the performance of different types of statistical algorithms using machine learning and classification techniques, combined with different balancing methods. The results indicated a significant improvement of accuracy and average error when the algorithms were applied in a balanced corpus, comparing with the unbalanced ones.