Adriano Del Pino Lino

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.

A Proposal for Automatic Evaluation by Symbolic Regression in Virtual Learning Environments

Empirically, symbolic regression tries to identify, through genetic programming and within the sphere of mathematical expressions, a model which best explains the relationship between variables in a given set of data, in terms of precision and simplicity. Virtual learning environments focused on evaluation have been previously investigated, as they offer teachers an effective teaching and learning tool and the student the possibility of computer-assisted evaluation and customized learning. Within this context, the present paper introduces an alternative approach to automatic evaluation in virtual learning environments, which offers the following improvements when compared to other methods, as superior accuracy when compared with the linear regression method, simplicity of implementation and context adaptive. To this extent, it presents the benefits of symbolic regression through genetic programming, emphasizing its efficiency and simplicity of implementation.

Assessing Review Reports of Scientific Articles: A Literature Review

Computational support has been applied in different stages for automation of the peer review process, such as reviewer assignment to the article, review of content of the scientific article, detection of plagiarism and bias, all applying Machine Learning (ML) techniques. However, there is a lack of studies which identify the instruments used to evaluate the reviewers’ reports. This systematic literature review aims to find evidence about which techniques have been applied in the assessment of the reviewers’ reports. Therefore, six online databases were evaluated, in which 55 articles were identified, all published since 2000, meeting the inclusion criteria of this review. The result shows 6 relevant studies, which address models of assessment of scientific article reviews. Nevertheless, the use of ML was not identified in any case. Therefore, our findings demonstrate that there are a few instruments used to assess the reviewers’ reports and furthermore, they cannot be reliably used to extensively automate the review process.

Educational data mining to track students performance on teaching learning environment LabSQL

Educational environments have added the software use to support teaching activities. In distance learning courses its common to store large amounts of data, which detail record the activities developed by the students. Those data can be used by professionals for the discovery of informations that can help teachers and students to manage and monitor the teaching-learning process. This work aims to obtain inferences related to the performance of the students, based on the data obtained from the learning enviroment, called Language Teaching Laboratory SQL(LabSQL), through the use of Educational Data Mining techniques (MDE). From the experiments done, we propose an architecture model for an interface based on educationals mining data in order to help students and teachers during the course of the discipline.

Criminal data mining: A case studyin Criminal Observatory Tapajós

Data mining allows research to reach patterns, often not visible by the simple observation of data. It is an exploration process in order to detect relations between the variables, seeking infer or create forecasts for future data. In public safety area data mining or MD, can be used in several ways: to identify the relation of a crime type with some neighborhood, determine the existence of a pattern for age, sex, day and the time that somebody commits some type of crime, among many other possibilities. The purpose of this article is to use MD technique in OBCRIT - Criminal Observatory Tapajós, which is a database that records reports of occurrences of the 3rd Battalion of Military Police of Pará State. The simulation results using Weka tool proposes to trace a profile of the occurrences that are part of the same group, finding common indications for the registered crimes and showing the importance of the use of data mining to in the process of extract knowledge to criminal levels.