What is it about?
This work divides the moodle data sets from six different sections of an annual postgraduate program at the Hellenic Open University in six periods for each section, due to the number of written assignments. Then it implements data mining techniques to analyze the activity, polarity and emotions of tutors and students in order to predict students’ grades. The results indicate the algorithm with the highest precision in each prediction. In addition, the research concludes that polarity and emotions as independent variables provide better performance in comparative models. Moreover, tutors' variables are highlighted as an important factor for more accurate predictions of student grades. Finally, a comparison of actual and predicted grades indicates which students have used a third party to fulfill their assignments.
Featured Image
Photo by Markus Spiske on Unsplash
Why is it important?
The main effort of this work is the implementation of Big Data methodologies to provide timely, personalized and as accurately as possible the scores of eight target classes. These are the six written assignments, the total average grades and the final grades of the students' assessment. The data origins are six different sections of an annual postgraduate program at the Hellenic Open University (HOU). By applying ML and developing regression algorithms, paper studies model behavior in six different periods of an academic year, adding variables from structured and unstructured LMS data. In addition to student interactions with the platform contents, we also look at the predictive value of students' raw data in forum interactions with their classmates, tutors and vice versa. To this end, a sentiment analysis is applied to characterize the polarity of students’ and tutors’ texts, to reveal their emotions and finally examine the involvement of the feeling in the precision of the models. As the state of emotions is not stable, it is examined in six different periods during the academic year. Polarity is represented as three categorical approaches and feelings as eight dimensional approaches. Furthermore, this paper studies the tutor's activity both in structured and raw data, analyzes their emotional state and examines their predictive contribution to the eight categories of student grade objectives. Moreover, a comparison of actual and predicted grades indicates the number of students, which of them and in what section a third party is used to fulfill their assignments.
Perspectives
Following the global trend in tertiary education, the current project implements ML and compares four models to predict each student's grades. Taking into account the current literature, this paper looks at three additional perspectives. In terms of the first, investigates the tutors’ activity through their authentic academic activity in the LMS as predictive factors in our models. In terms of the second, apart from students, the tutor's sentiments are also studied. Forum texts are used to describe polarity and emotions in three and eight characteristics respectively for both concerned. The analysis of tutors' polarity and emotions is not based on students' views on academic work or tutor support, as some are not objective due to emotional load but are derived from DM methodology [40]. As far as the last perspective is concerned, tutors' and students' moodle activity blend with their polarity and emotional states as prognostic variables for 8 different target classes. The methodology of the project also leads to a comparison of predictive and actual grades, indicating students who were most likely addressed to third parties to complete their written assignments. Timely prediction of student grades in different periods during the academic year, taking into account, the LMS typical activity and the changing emotional state of platform’s daily users, promotes tutors’ self-evaluation and focused interventions, learning content evolution but mainly, students’ retention in the learning process.
Dr Andreas Fotios Gontzis
University of Patras
Read the Original
This page is a summary of: Polarity, emotions and online activity of students and tutors as features in predicting grades, Intelligent Decision Technologies, September 2020, IOS Press,
DOI: 10.3233/idt-190137.
You can read the full text:
Resources
Sentiment Analysis to Track Emotion and Polarity in Student Fora
The purpose of this paper is to propose a data mining methodology for analysing data relating to the participation of students in the online forum of a postgraduate course at the Hellenic Open University. Data is migrated to MongoDB, a NoSQL database management system, and analysed using the rmongodb package of R statistical environment. We focus in sentiment analysis to extract the emotional knowledge of students' fora. Polarity and emotion are identified in messages and are classified as positive, negative or neutral. Messages are categorized and visualized in six basic emotions, as a multiclass approach in understanding students' written opinion. By identifying sentiment behaviour from students' discussion fora, we are able to assess the effectiveness of the learning environment to improve students' learning experience, tutors' instructional experience and the university's institutional strategic view.
A Holistic View on Academic Wide Data through Learning Analytics Dashboards
We live in an era of ever developing technology which has led to a massive increase in the amount of data available. Every move, click, and swipe of a card creates a virtual image of our lives in the form of a personal mosaic. Through sophisticated methods applied on these data, companies are now able to predict whether their products will appeal to people and target their advertising and market with maximum profit. In thecompetitive and globalized environment of education, institutions have to attract, assess and guide their students. In Greece, the Hellenic Open University (HOU) offers its courses in an open and distance learning mode. In contrast to a traditional university where most of the interaction and teaching is taking place on a face-to-face basis, at HOU the learning processis mainly facilitated via multimodal technological pathways and systems. It is important to homogenize and integrate the data collected from these systems in order to utilize it by gaining knowledge and building on it. To accomplish this, we use a learning analytics methodology to analyze the data and automatically create a detailed and holistic image of student performance, tutor effectiveness, and administration efficiency. This is then visualized through learning dashboards which convey important information so that each party can take necessary action and help the institution to improve its standing in the competitive educational environment. We aim to use the information within the systems where it is derived, since this makes the process more user friendly and accessible to all those involved.
A Predictive Analytics Framework as a Countermeasure for Attrition of Students
Attrition is one of the main concerns in distance learning due to the impact on the incomes and institutions reputation. Timely identification of students at risk has high practical value in effective students’ retention services. Big Data mining and machine learning methods are applied to manipulate, analyze, and predict students’ failure, supporting self-directed learning. Despite the extensive application of data mining to education, the imbalance problem in minority classes of students’ attrition is often overlooked in conventional models. This document proposes a large data frame using the Hadoop ecosystem and the application of machine learning techniques to different datasets of an academic year at the Hellenic Open University. Datasets were divided into 35 weeks; 32 classifiers were created, compared and statistically analyzed to address the minority classes’ imbalance of student’s failure. The algorithms MetaCost-SMO and C4.5 provide the most accurate performance for each target class. Early predictions of timeframes determine a remarkable performance, while the importance of written assignments and specific quizzes is noticeable. The models’ performance in any week is exploited by developing a prediction tool for student attrition, contributing to timely and personalized intervention.
Measuring Engagement to Assess Performance of Students in Distance Learning
Organizations continuously invest in the analysis of accumulated data by linking their exploitation to more effective decision making. Higher education and especially distance learning as a large data tank, follows the technological and financial needs to involve more flexible data analysis environments and better data-informed decision capabilities. Shrinking public subsidies drives higher education to form a more competitive learning environment. Satisfactory user experience and personalized services require a quantitative and qualitative analysis of students' daily action in learning environments. The increasing adoption of Learning Analytics (LA) and Educational Data Mining (EDM) push the development of novel approaches and advancements in education sector. At the same time, the rapid disclosure of hidden knowledge and the immediate presentation of results to optimize personalized decision making is a challenge for the competitiveness of distance learning. The objective of this study is the analysis of processes, technologies and resources used in an annual module at the Hellenic Open University to provide stakeholders with the visibility of interactions and hidden value in students' interactions. LA technologies are used on large set of data that has been collected by a Moodle platform and carefully analyzed. Students' logins, replies and quizzes are blended with the average grade of the main written exercises during the academic year. The contribution of this work is the useful observations from the students' educational on-line activities as predictive factors for their academic performance.
Contributors
The following have contributed to this page
