Relatório de Artigos - Included

Resumo: This paper examines the use of Educational Data Mining (EDM) to predict the academic performance of elementary students specifically in Mathematics. It explores ten Machine Learning classifiers, comprising eight base learners (Linear SVM, Logistic Regression, Medium KNN, Wide NN, Fine Decision Trees, Bilayered NN, Fine KNN, and Medium NN) as well as two ensemble learners (Ensemble Subspace Discriminant and Ensemble Boosted Trees) within the MATLAB environment. The analysis utilizes a dataset featuring 33 academic and demographic features of 280 students. To mitigate the imbalanced distribution in class data, resampling techniques such as Random Under-Sampling Boost (RUSBoost), Synthetic Minority Oversampling Technique (SMOTE), and hybrid combinations of both are employed. The experimental outcomes demonstrate that the hybrid-sampling SMOTERUSBoosted Trees algorithm achieves the highest accuracy of 75% on testing data, indicating the efficacy of combining oversampling and under-sampling techniques for modeling imbalanced datasets. This finding underscores the potential of EDM in the elementary education sphere to bolster data-driven interventions and enhance students’ Mathematics achievement.

Resumo: This study examines whether a multi-class predictive machine learning model can be used to predict students’ math performance in computer-based international large-scale assessments. Raw data from the PISA 2012 computer-based assessment was used to conduct the experiments. Three-class neural network and random forest machine learning models were developed and compared in prediction performance for students’ proficiency in mathematics. The three levels of math proficiency were low, mediocre and high. Both machine learning models relied on background variables from student, parental and school questionnaires, as well as on the student math assessment data. Manual and a few automatic feature selection methods were explored while developing the machine learning models. Due to the imbalance in the three-class dataset Cohen’s Kappa coefficient and ROC AUC were used as main metrics of the model performance.

Resumo: Education is crucial for the development of any country. Analysis of education datasets requires effective algorithms to extract hidden information and gain the fruitful results to improve academic performance. Multiple models were used to maximize the contribution to the education environment. In this study, we used the spot-checking algorithm to compare these methods and find the most effective method. We propose three main classes of education research tools: a statistical analysis method, machine learning algorithms, and a deep learning framework. The data were obtained from many high schools in Cambodia. We introduced feature selection techniques to figure out the informative features that affect the future performance of students in mathematics. The proposed ensemble methods of tree-based classifiers provide satisfiying results, and in that, random forest algorithm generates the highest accuracy and the lowest predictive mean squared error, thus showing potential in this prediction and classification problem. The results from this work can be used as recipe and recommendation for mining various material settings in improving high school student performance in Cambodia.

Resumo: Educational Data Mining is a field of study that aims to find patterns and information in educational institutions through mining educational data. To become a better teacher, teachers need to anticipate their pupils' performance patterns. Knowledge gained from it can be used in various ways, such as a strategic plan for delivering high-quality education. This report suggests that students' final grades can be predicted using data mining techniques based on past research. On two educational datasets related to mathematics classes and Portuguese language lessons, three well-known data mining approaches, such as Decision Tree, JRip, Naive Bayes, Multilayer Perceptron, and Random Forest, were utilized in the experiments. As a result, using the employed data mining methods, student success might be predicted with reasonable accuracy.

Resumo: This study focuses on designing a personalized learning path optimization algorithm based on deep learning, aiming to improve learning efficiency and quality. The algorithm innovatively integrates knowledge graph and reinforcement learning, and introduces sentiment analysis. The knowledge graph clearly presents the course knowledge system, reinforcement learning dynamically adjusts the path according to the real-time status of students, and sentiment analysis takes students' emotions into consideration to optimize the learning experience. In the experimental stage, the performance data of multiple courses (such as mathematics, Chinese, and English) in one semester were collected, including regular homework, quizzes, and final exam scores, and students' learning behavior logs were recorded. A comparison was made between traditional collaborative filtering and shallow neural networks based on deep learning recommendation. Experimental results showed that compared with control group, the learning effect was significantly improved by 15%, learning time decreased by 20%, and user satisfaction was 4.2 (5 points). This fully verifies the excellent performance of the new algorithm in generating personalized learning paths, and provides strong support for the development of personalized learning in the field of education.

Resumo: Abstract This paper first proposes the direction of constructing a higher mathematics teaching mode supported by intelligent technology and then models the learner portrait of the learning outcome data according to the subject knowledge graph. An improved ant colony optimization algorithm is used to search for the optimal learning path, which is then combined with an improved convolutional neural network to generate a personalized learning path. The Trans R method is used to quantify the relationship between learners and learning resources and a semi-supervised learning conditional random field method based on K-NN is proposed to label learning resources and generate learning accurate evaluation for smart teaching. The smart teaching model of advanced mathematics courses is applied and analyzed in terms of students’ advanced mathematics pre and post-test scores, students’ satisfaction, teachers’ teaching methods and teaching resources in four directions. The analysis obtained that the posttest scores of the learners in the experimental group were 75.631, and the posttest scores of the learners in the control group were 66.314, with a difference of 9.317. The significance level of the variance chi-square test was 0.000<0.05, which shows that there is a significant difference between the posttest scores of the experimental group and those of the control group, and it indicates that the wisdom teaching of higher mathematics significantly enhances the learning performance of the learners.

Resumo: Educational data mining (EDM) is a rapidly growing research area, and the outputs obtained from EDM shed light on educators' and education planners' efforts to make efficient decisions concerning educational strategies. However, a lack of work still exists on using EDM methods for international assessment studies such as the International Association for the Evaluation of Educational Achievement's Trends in International Mathematics and Science Study (IEA's TIMSS). This study aims to fill the gap in the current literature on the latest-released TIMSS 2011 data by applying a decision tree, a Bayesian network, a logistic regression, and neural networks. The best performing algorithm in classification based on several performance measures has been found for eighth-grade Turkish students' mathematics data. During the construction of models, 11 student-based factors have been taken into account. The results show that logistic regression outperforms other algorithms in terms of measuring classification performance. The factor of student confidence has also been found as the most effective factor on eighth-grade students' mathematics achievement.

Resumo: Machine Learning (ML) is expected, in the near future, to provide various venues and effective tools to improve education in general, and Science-Technology-Engineering- Mathematics (STEM) education in particular. The Gartner Analytics Ascendancy Model requires the use of four types of data analytics to be considered comprehensive: descriptive, diagnostic, predictive and prescriptive data analytics. This paper presents the outcomes of a research and development project at Bradley University (Peoria, IL, USA) aimed at the setup and benchmarking of eight ML algorithms for predictive learning analytics, specifically, a prediction of student academic performance in a course. The analyzed and tested ML algorithms include linear regression, logistic regression, k- nearest neighbor classification, naïve Bayes classification, artificial neural network regression and classification, decision tree classification, random forest classification, and support vector machine classification. Based on the obtained accuracy of the analyzed and tested ML algorithms, we have formulated a set of recommendations for faculty and practitioners in terms of selection, setup and utilization of ML algorithms in predictive analytics in STEM education. We also performed formative and summative surveys of undergraduate and graduate students in Computer Science and Computer Information Systems courses to understand their opinion about utilization of ML-based predictive analytics in education; a summary of obtained student feedback is presented in this paper.

Resumo: Intelligent tutoring systems are effective for improving students’ learning outcomes (Bowen et al., 2013; Koedinger & Anderson, 1997; Pane et al., 2013). However, constructing tutoring systems that are pedagogically effective has been widely recognized as a challenging problem (Murray, 1999, 2003). In this thesis, I explore the use of computational models of apprentice learning, or computer models that learn interactively from examples and feedback, to support tutor development. In particular, I investigate their use for authoring expert-models via demonstrations and feedback (Matsuda et al., 2014), predicting student behavior within tutors (VanLehn et al., 1994), and for testing alternative learning theories (MacLellan, Harpstead, Patel, & Koedinger, 2016). To support these investigations, I present the Apprentice Learner Architecture, which posits the types of knowledge, performance, and learning components needed for apprentice learning and enables the generation and testing of alternative models. I use this architecture to create two models: the DECISION TREE model, which non- incrementally learns when to apply its skills, and the TRESTLE model, which instead learns incrementally. Both models both draw on the same small set of prior knowledge for all simulations (six operators and three types of relational knowledge). Despite their limited prior knowledge, I demonstrate their use for efficiently authoring a novel experimental design tutor and show that they are capable of achieving human-level performance in seven additional tutoring systems that teach a wide range of knowledge types (associations, categories, and skills) across multiple domains (language, math, engineering, and science). I show that the models are capable of predicting which versions of a fraction arithmetic and box and arrows tutors are more effective for human students’ learning. Further, I use a mixedeffects regression analysis to evaluate the fit of the models to the available human data and show that across all seven domains the TRESTLE model better fits the human data than the DECISION TREE model, supporting the theory that humans learn the conditions under which skills apply incrementally, rather than non-incrementally as prior work has suggested (Li, 2013; Matsuda et al., 2009). This work lays the foundation for the development of a Model Human Learner— similar to Card, Moran, and Newell’s (1986) Model Human Processor—that encapsulates psychological and learning science findings in a format that researchers and instructional designers can use to create effective tutoring systems.

Resumo: AbstractSeeing mathematics teaching as a very demanding and responsible process while having in mind the importance of mathematical knowledge for students of technical faculties, this paper aims to present heuristics for student classification according to their predicted mathematical success. Over the last few decades, the process of informatization of universities has resulted in new challenges universities are faced with. Due to the widespread use of educational databases, which opens new possibilities for educational data mining and analyses, machine learning algorithms have become a very popular tool for predicting students' academic performance. The decision tree algorithm is used in this paper for the classification and prediction of students' mathematical performance and it is trained on the data collected from the educational information system. The experimental results show that the model accuracy is 72% with an error rate of 0.28. The implementation of the Decision Tree Model to predict whether a student will pass, fail or be conditional in mathematical courses is important for both teachers and students, as well as for universities. Students' performance is one of the major keys in evaluating the quality of the teaching process, but also for evaluating the overall success of the university itself. As mathematics is considered a basic and important discipline, it is clear why predicting students' mathematical achievement is crucial for all levels of university organization.

Resumo: An essential component of the educational activity is rigorous examination and assessment of students' results, with a potential substantial influence on student growth. This paper offers a predictional model for student's success in secondary education using five classification algorithms: Logistic Regression, K-Nearest Neighbors (KNN), Support Vector Machine (SVM), XGBoost, and Naive Bayes, where the data is gathered from two Portuguese school reports and surveys. The two core disciplines (Mathematics subject and Portuguese language) of the dataset were modeled around binary/five-level classification tasks which is imbalanced. The imbalanced dataset is also balanced by using K-Means SMOTE (Synthetic Minority Oversampling Technique) before classification. The test results reveal that obtaining the most outstanding accuracy value is 96.89% of Support Vector Machine (SVM) is superior to Logistic Regression, KNN, XG-Boost, and Naive Bayes. Therefore, it is essential to consider whether a model makes a particular prediction. Thus, we then train an interpretable LIME (Local Interpretable Model-agnostic Explanations) model for all the classifiers and the construction of explainable models can have a major advantage: the model can be confident, the transparency of the model helps to understand the underlying processes for working.

Resumo: The ability to predict the performance tendency of students is very important to improve their teaching skills. It has become a valuable knowledge that can be used for different purposes; for example, a strategic plan can be applied for the development of a quality education. This paper proposes the application of data mining techniques to predict the final grades of students based on their historical data. In the experimental studies, three well-known data mining techniques (decision tree, random forest, and naive Bayes) were employed on two educational datasets related to mathematics lesson and Portuguese language lesson. The results showed the effectiveness of data mining learning techniques when predicting the performances of students.

Resumo: Purpose – Geometry plays a crucial role in developing cognitive skills such as spatial reasoning, visualization, and problem-solving. However, many students in Ghanaian senior high schools face difficulties with abstract topics like circle theorems. This study examines the effectiveness of combining jigsaw cooperative learning with computer animation to improve students’ conceptual understanding of geometric concepts compared to traditional teaching methods. Methodology – A quasi-experimental design was adopted involving senior high school students assigned to control and experimental groups. The control group received conventional instruction, while the experimental group was taught using jigsaw cooperative learning supported by computer animations. Pre-test and post-test data were collected and analyzed using the Mann-Whitney U test due to non-normal data distribution. Findings– Students in the experimental group significantly outperformed those in the control group, demonstrating higher post-test scores. The integration of cooperative learning and visual animation enhanced conceptual understanding, reduced cognitive load, and improved knowledge retention. Novelty – This study offers a unique contribution by integrating jigsaw learning with computer animation—a combination rarely explored in teaching abstract geometry. Conducted in a sub-Saharan African context, it extends limited research on multimedia-supported instruction by focusing not only on academic performance but also on deeper cognitive outcomes. Significance – The findings underscore the potential of technology-enhanced collaborative strategies in improving learning in abstract mathematical domains. The study provides evidence-based recommendations for adopting innovative pedagogies in low-resource educational settings, with implications for curriculum development and teacher training.

Resumo: This paper derives from a large research project focusing on mathematics and science assessment of student learning in three high-need, rural, and urban secondary schools in Manitoba, Canada. The study employed qualitative methods of semi-structured interviews and classroom video recordings of teaching practice experiences of 12 mathematics and science teachers, with the purpose that explore how authentic assessment forms assist effective teaching to monitor and motivate student learning achievement and growth. The results indicate that about 67% (eight out of the twelve of the participants) of the research participants practice the traditional mode of standard assessment that consists of multiple forms of questioning. The participants' rationale relates to speedy evaluations of student work, preparing feedback reports to parents and students, and objectivity of the assessment process. The other 33% (four out of twelve of the participants) of participants practice authentic assessment that concentrates on: (1) Allowing students to apply what they have learned rather than testing their ability to memorize and regurgitate concepts, (2) Allowing students to personalize their knowledge and values, (3) Encouraging group project-based learning and with the use of rubric for evaluating and monitoring, (4) Promoting deep learning to become life-long learners, (5) Recognizing, acknowledging, and validating diversity in student learning styles, interests, and aspirations, and further, authentic assessment is an excellent opportunity to apply communicative technologies such as podcasts and webinars in learning and undertaking investigations in mathematics and science learning. Furthermore, some participants asserted that authentic assessments are time-consuming, labor-intensive, and resource-demanding, aside from the limited resources and lack of training, which are some of the challenges of implementing authentic assessment. Other participants stated that all teachers must be familiar with using all assessment tools. The paper concludes that the principal plays a critical instructional leadership role in a school-wide implementation of authentic assessment.

Resumo: This study aims to examine the improvement of metacognitive abilities in learning mathematics through a variety of formative assessments. The study applied an experimental approach with a pre- and post-tests control group design. Six classes of students were chosen by cluster random sampling   as the sample, with two classes serve as the experimental group with performance assessments, two classes serve   as the comparison group with essay assessments   and two additional classes serve as the control group with multiple choice assessments.  The instrument to measure metacognitive ability was developed specifically for the pre-and post-tests. The gain score was analyzed using a one-way ANOVA and continued with the Scheffe test. The study discovered that students who participated in learning with formative assessments of performance showed the highest levels of metacognitive skills followed by those who participated in learning with formative assessments of essays and those who participated in learning with formative assessments of multiple-choice questions came in third. These findings lead to the conclusion that formative assessment of performance has a positive effect on improving metacognitive abilities. According to the characteristics of learning mathematics, this situation might happen because the performance evaluation includes activities that are difficult, comprehensive and associated with daily life.   In addition, performance assessment also has intrinsic value because it requires students to organize and present material in their own way. It is recommended that mathematics teachers use formative performance assessments in order to enhance cognitive capacities.

Resumo: Adaptive learning systems have emerged as a transformative approach in K-12 education, promising to personalize learning experiences and improve educational outcomes. This paper aims to assess the effectiveness of these systems in enhancing student learning, engagement, and achievement by focusing on their integration in various K-12 settings and examining both quantitative and qualitative outcomes. Adaptive learning systems utilize datadriven algorithms to tailor educational content to the individual needs of students, providing personalized learning paths that adapt in real-time based on student performance and engagement. These systems are designed to address diverse learning styles, paces, and abilities, potentially closing educational gaps and promoting equity in learning. The primary objectives of this study are to evaluate the impact of adaptive learning systems on student achievement in core subjects such as mathematics, science, and reading; to assess changes in student engagement and motivation resulting from the use of adaptive learning technologies; and to identify best practices and challenges associated with the implementation of adaptive learning systems in K-12 classrooms. A mixed-methods approach was employed to provide a comprehensive evaluation, combining a quasi-experimental design with control and treatment groups across multiple schools for the quantitative component, and surveys, interviews, and classroom observations for the qualitative component. Quantitative data included pre- and post-tests, standardized test scores, and attendance records to measure academic achievement and engagement levels, while qualitative data provided insights from students, teachers, and administrators regarding their experiences with adaptive learning technologies. The study found significant improvements in student achievement in schools that implemented adaptive learning systems, with students demonstrating higher gains in test scores compared to their peers in control groups. The personalized nature of adaptive learning helped struggling students catch up with their peers, particularly in mathematics and reading. Additionally, students reported increased engagement and motivation, attributing their interest to the interactive and tailored learning experiences provided by adaptive technologies. Teachers noted that adaptive learning systems facilitated differentiated instruction, allowing them to cater to individual student needs more effectively. However, challenges such as initial setup costs, the need for ongoing professional development, and varying levels of technology access were identified. Successful implementation was closely tied to strong support from school leadership and sufficient training for educators. The findings suggest that adaptive learning systems hold considerable promise for enhancing educational outcomes in K-12 settings. The ability of these systems to provide real-time feedback and personalized learning paths can significantly benefit students, particularly those who require additional support. However, the effectiveness of these systems is contingent upon proper implementation, including adequate training for teachers and equitable access to technology for all students. The study highlights the importance of addressing the digital divide to ensure that all students can benefit from adaptive learning technologies. Future research should explore long-term impacts, including the sustainability of academic gains and the potential for adaptive learning systems to foster lifelong learning skills. Adaptive learning systems have demonstrated their potential to transform K-12 education by personalizing learning experiences and improving student outcomes.

Resumo: The aim of this quantitative research is to analyze the impact of Facebook in the teaching-learning process in financial mathematics education, using data science, machine learning, and neural networks. The sample is composed of 46 students from the Bachelor of Administration, Commerce and Marketing program at La Salle University. The results of machine learning (linear regression) indicate that sending messages, watching instructional videos, and publishing exercises on Facebook supports the teaching-learning process in financial mathematics. Likewise, data science identified six predictive models for the use of Facebook in the educational context, by means of the decision tree technique. Analysis using neural networks identified the influence of sending messages, watching instructional videos, and publishing exercises on Facebook during the assimilation of knowledge and development of mathematical skills. Finally, Facebook is a technological and communication tool that transforms the organization of teaching and learning activities in financial mathematics education.