[Research Contribution] Analysis of Student Feedback in terms of Using Machine Learning Approach

11 March, 2025

Keywords: Student feedback analysis, Machine learning, Natural Language Processing, Artificial Intelligence, Smart Education

Student feedback is a crucial data source for educational institutions to improve the quality of teaching and learning environments. However, with the large volume of data and the diversity of natural languages, traditional feedback analysis has many limitations. In this context, Machine Learning has emerged as a potential solution for automating the classification, evaluation, and extraction of important information from student feedback. In this study, therefore, the researchers from the University of Economics Ho Chi Minh City – Vinh Long Campus (UEH Mekong) have suggested using machine learning approaches to analyzing domestic and international student feedback, identifying current trends and limitations to make timely recommendations for improvement and orientation of further research in this field.

Research context

In today’s era, higher education plays both an economic role and the role of training highly qualified human resources for other economic fields; in addition, this is also the center of innovation. In a dynamic educational environment, effective student feedback assessment is the basis for ongoing improvement and enhancement of the overall learning experience.

Nevertheless, the fact that traditional feedback analysis methods fail to provide timely detailed information and are highly subjective is also an inherent characteristic of manual assessment. The effectiveness of analyses of students’ opinions and thoughts will not be satisfied because of an overwhelming data source. Therefore, the development of Artificial Intelligence (AI), especially machine learning, has opened up new approaches to Natural Language Processing (NLP). Thanks to that, systems can automatically identify topics, classify emotions, and extract important information from student feedback. This helps educational institutions make accurate decisions, which can improve the quality of teaching and learning experiences.

In that context, the research from UEH Mekong clarifies the trends of interest in this field; concurrently, pointing out the limitations and gaps that still exist in current research. Afterthat, specific recommendations are offered for the development and the application of student feedback analysis using machine learning in Vietnam in the future.

Student Feedback and Machine Learning in Education

Student feedback includes opinions and assessments of quality, teaching methods, course content and learning experiences. While traditional analytics rely heavily on statistics, Machine Learning allows for the exploitation of open feedback data through Natural Language Processing (NLP), which automatically classifies, identifies topics, and measures students’ emotional levels towards each aspect of teaching.

Machine Learning (ML) is a branch of Artificial Intelligence (AI), which enables computers to learn from data and to make predictions or decisions without being explicitly programmed. Instead of following fixed rules set by humans, ML models improve themselves through experience and input data.

Natural Language Processing (NLP) supports a variety of tasks, allowing for keyword extraction, topic identification, and emotional analysis in student feedback. Deep Learning uses neural networks as a method in AI that teaches computers to process data in a way inspired by the human brain, using neural networks to understand context, and to detect general trends from student feedback.

Modern research trends in student feedback analysis

One of the important trends in student feedback analysis research is text classification, which aims to group feedback comments according to specific criteria, providing useful information for improving teaching quality. Of which, sentiment analysis is an important task, focusing on determining the emotional state (positive, negative, neutral) of students.

Currently, most studies perform sentiment analysis at the sentence or document level, aiming to evaluate students’ overall feelings concerning the learning experience. In addition, the new trend is expanding the scope of research, not only at general assessment but also delving into analysis of learning attitudes, teaching quality, and course content. Exploiting this feedback in depth assists educational institutions in improving the learning experience, optimizing teaching methods, and proposing policies that suit students’ needs.

In addition, another notable trend is the preference for supervised ML algorithms in analyzing student feedback. This reflects the need in the education sector for analytical models that return accurate and clear results, which is difficult for unsupervised learning algorithms to achieve and is the driving force behind the use of supervised learning models that operate on pre-labeled datasets to optimize the learning process and to improve prediction capabilities. Among them, shallow learning algorithms are the most popular because of their simplicity. In parallel, some pioneering studies have exploited the potential of deep learning models, including Long Short-Term Memory (LSTM), Convolutional Neural Network (CNN) and Multi-layer Perceptron (MLP) to improve the ability to analyze and to recognize the context of student feedback. Despite the diverse methods, data sources used in research mainly focus on responses from surveys. Survey responses have the advantage of being easy to control and highly reliable for ensuring data consistency, improving the accuracy of the analytical model. On the other hand, this means that studies are limited in exploiting richer data sources like responses from online learning platforms, social networks, or other non-traditional sources.

Limitations of student feedback analysis

First, many analytical models do not effectively handle special characters like emoticons or abbreviations – important factors reflecting students’ attitudes and emotions. Ignoring these factors reduces the accuracy of interpreting student feedback, affecting the quality of analysis.

Second, most current research uses supervised machine learning models, which rely on manually labeled data in a specific language or context. This makes it difficult to apply to other environments because of the limitations in scalability and practical applicability.

Third, most research only focuses on model accuracy without paying attention to other important factors like data imbalance resistance and response time optimization. This is an incomplete approach in current research.

Policy implications for student feedback analysis in Vietnam

From the current domestic and international literature, some important policy implications are offered to improve the effectiveness of student feedback analysis in Vietnam:

First, there should be more investment in applying support techniques and preprocessing of Vietnamese feedback, including detecting and classifying sarcastic labels and comment spam; and applications of text feature analysis models such as punctuation and emotional changes to interpret students’ attitudes more clearly so as to improve the accuracy of the analysis model.

Second, the level of detail in sentiment analysis should be increased, focusing on specific factors related to education. In particular, aspect-based sentiment analysis is an important task requiring the support of Named Entity Recognition (NER) models to extract relevant factors in the feedback. Because of the specificity of the Vietnamese language, these models need to be developed based on student feedback data to improve accuracy.

Third, the student feedback dataset needs to be expanded and standardized. UIT-VFSC is currently the only standard dataset on student feedback in Vietnam. However, it is necessary to supplement data to balance low-frequency labels and to build datasets from online sources and learning platforms to increase the diversity and the practicality of analytical models.

Fourth, it is necessary to optimize the model to overcome data imbalance and to optimize response time. Using pre-trained models can help Vietnam reduce dependence on original data while limiting the negative impact of data imbalance. For response time optimization, the model pruning method can remove less important parameters, reduce the computational load while maintaining the required accuracy.

The study demonstrated that the application of machine learning in student feedback analysis is an important trend, with two main focuses being text classification and sentiment analysis. Currently, supervised machine learning algorithms are being popular because of their high accuracy; on the contrary, there exists some difficulty applying them to multi-context environments. Meanwhile, deep learning models promise the potential for in-depth analysis but have not been widely deployed because of high requirements for data and computational resources.

Based on international research trends, the study suggested the following solutions: sarcasm detection, special character processing, aspect-based sentiment analysis, and data set expansion. These directions will help educational institutions use the analysis results more effectively to improve teaching quality.

In general, machine learning plays an important role in supporting educational institutions to analyze feedback accurately and comprehensively, improving the learning experience of students. For effective application, educational institutions need to invest in natural language processing systems and develop AI models suitable for the characteristics of Vietnamese. Combining this technology not only optimizes the educational management process but also creates a modern learning environment better meeting the needs and expectations of students.

The full-text research article Analysis of student feedback using machine learning approach can be accessed HERE.

Authors: Tran Son Nam, Nguyen Nha Yen – University of Economics Ho Chi Minh City

This article is part of a series spreading research and applied knowledge with the message “For a More Sustainable Mekong”, under the program “Research Contribution For All” implemented by UEH. UEH cordially invites readers to read the next UEH Research Insights newsletter.

News, photos: The authors, UEH Mekong Department of Admissions and Communications, UEH Department of Communications and Partnerships