Virtual teaching and learning environments: automatic evaluation with artificial neural networks

Forecasting techniques have been widely used in automatic assessment in virtual teaching and learning environments, allowing educators to make decisions in both present and future planning. However, due to the element of uncertainty intrinsic to the forecasting methods, a number of studies have been carried out in order to find a more efficient model that allows for exploring and inferring the relation of a dependent variable with independent variables. In this context, we propose an alternative to solve the problem of automatic evaluation with the use of artificial neural networks that are adjusted, or trained, so that a certain input leads to a specific target output. Therefore, the research seeks to achieve the following: (1) review the state-of-the-art work published in this area, (2) propose a better forecasting model as compared to the existing model, (3) perform a comparative analysis with the previous experiments of multiple linear regression (MLR) and symbolic regression, (4) propose future research guidelines. To this end, a case study was applied to clarify the benefits of the artificial neural networks, emphasizing its efficiency and simplicity of implementation. With a margin of error of less than 2%, this proposal simulates a specialist, and automatically evaluates a student’s answer. As a result, the proposed model performance overcomes the methods of multiple linear regression and symbolic regression efficiently, eliminating the problem of randomness, reducing processing time and at the same time providing a model with higher accuracy and lower error rates.