TF-IDF Method Research Articles

Modeling of Naïve Bayes and Decision Tree Algorithms to Analyze Sentiment Related to Jaklingko Public Transportation on Social Media X (Twitter)

Technological developments have changed the way people access information and share opinions through the internet and social media such as X (Twitter). Public sentiment is now crucial in evaluating public services, especially in the era of advanced information technology. As a metropolitan city, public transportation in DKI Jakarta plays an important role in economic, business and government activities. The Jaklingko initiative launched by the DKI Jakarta Provincial Government aims to provide efficient and affordable public transportation. This research implements and compares Naive Bayes and Decision Tree classification methods to perform sentiment analysis of twitter users opinions regarding Jaklingko into positive and negative categories. Data was collected by crawling using tweet_harvest which obtained 6001 tweets about Jaklingko, then text preprocessing was carried out. Word weighting is done using the TF-IDF method to give value to each term in the document, and sentiment labeling is done using the Vader Lexicon library. The data is divided into training and testing data with a ratio of 80%:20% for the classification process. Evaluation of the method is done using confusion matrix. The results showed that the accuracy of Naive Bayes reached 84.9% and Decision Tree reached 84.2%. The wordcloud visualization depicts negative words including vehicle stoppage, bad driver attitude, and envy from people in other cities. Meanwhile, positive words included free system, useful programs, and user convenience. This research provides an in-depth understanding of public opinion towards Jaklingko, with potential implications for improving public transportation services in Jakarta.

A text-based recommender system for recommending relevant news articles

Despite recent advances in the related fields and the growing popularity of AI-based tools, small businesses, and public institutions still face challenges when implementing recommendation systems to increase profits and provide personalised services. This paper provides a comprehensive overview of state-of-the-art systems and a case study of a recommender system for a small, low-budget project, battling with several constraints: the use of uncommon natural language, the use of raw, unlabelled textual data, thus using mainly techniques coming from the field of data mining and text mining rather than relying on supervised learning methods. Another constraint was the reliance on the smaller dataset and the limited resources available for the development. This led us to explore a range of content-based methods, including the utilisation of the similarity measures applied to word embeddings derived from the Word2Vec and Doc2Vec shallow neural network models, as well as the TF-–IDF method. Additionally a topic modelling approach utilising the Latent Dirichlet Allocation was used, as well as collaborative filtering methods, such as the algorithm using the Singular Value Decomposition method, and hybrid methods integrating the fuzzy inference and fuzzy expert systems. Some of the obstacles identified were subsequently demonstrated to be too challenging for the development of accurate recommendations. However, the item-to-item similarity solutions, which were primarily content-based, yielded satisfactory results when the threshold of 75% of the average precision of recommendations assessed by users was exceeded. One such solution was the one that used the Word2Vec-based model, which had been trained on the parameters obtained from the word similarities and analogies tests. Furthermore, an overview of alternative techniques and methodologies is provided.

Sentiment analysis with machine learning for drug reviews

In the treatment of the disease, the fact that individuals use drugs independently from doctors without appropriate consultation causes their health status to become worse than normal. This article aims to conduct a sentiment analysis over the comments of individuals about the drug in case they use drugs without consultation. Within the scope of this study, patients' comments about drugs were vectorized using Bow and TF-IDF algorithms, sentiment analysis was made, and the predicted sentiments were; it was evaluated with precision, recall, f1score, accuracy and AUC score. As a result of the evaluations, the most successful result was obtained in the TF-IDF method. This result is the result of the Linear Support Vector Classifier algorithm with an Accuracy value of 93%.

GazeGCN: Gaze-aware Graph Convolutional Network for Text Classification

Graph convolutional networks (GCNs) are capable of capturing contextual relationships in text classification. In this paper, we propose a novel Gaze-aware Graph Convolutional Network (GazeGCN) for text classification, where the gaze signals from humans are incorporated into the GCN architecture, empowering our GazeGCN to use gaze information that can be shared across all data to model the intricate relationships between words and documents. To be specific, we first build a gaze prediction model to obtain five gaze signals to form the gaze distribution of each word. Then, Wasserstein Distance is employed to derive the gaze-aware word–word weight by calculating the gaze distribution between words, so as to improve the capture of luxuriant and similar relationships between words. Furthermore, to enhance the extraction of contextual syntactic information, we introduce a Gaze-enhanced TF–IDF method integrating gaze signals and term frequency–inverse document frequency (TF–IDF) for gaze-aware word-document weight derivation, thus making up for the lack of TF–IDF that does not consider syntactic information. Afterward, the rich source of graph edge information including gaze-aware word–word weight and gaze-aware word-document weight is incorporated to construct graphs, so as to leverage the relationship of word–word and word-document. Experiments show encouraging results on seven benchmark datasets that our approach outperforms the state-of-the-art baseline methods.

Smart Hostel Allocation System Using Naïve Bayes Algorithm

This research explores the potential of automated personality classification using supervised machine learning techniques, with a specific focus on the Myers-Briggs Type Indicator (MBTI) for hostel allocation. The problem is the manual and resource-intensive allocation of hostel accommodations in educational institutions, leading to personality mismatches among roommates, which negatively impacts students' academic engagement and well-being. The aim of this research is to improve hostel room allocation by predicting student personality traits, thus minimizing personality mismatches among roommates. A Naïve Bayes classifier was applied to the MBTI personality dataset acquired from Kaggle. Various preprocessing techniques, such as tokenization, stop-word removal, and word stemming, were used to prepare the dataset for feature selection. CountVectorizer and TF-IDF methods were employed to convert text into numerical vectors for machine learning analysis. Additionally, SMOTE was implemented to address class imbalance within the dataset, ensuring better model performance. The Naïve Bayes classifier showed promising results, with scores achieved across all personality traits as 73% precision, 70% accuracy, F1 Score 70% and Recall 70% in predicting MBTI personality traits from input text. Visualizations and performance metrics such as accuracy, precision, recall, and F1-score were used to evaluate the model's efficiency. The research findings suggest that supervised machine learning techniques can effectively predict personality traits from social media data, aiding in improved hostel room allocation and potentially benefiting broader applications in education and computational psychology. This system is recommended for educational institutions and residential complexes seeking efficient roommate pairing.

Natural Language Processing Methods Application in Defense Budget Analysis

Abstract Transferring to economy 5.0 makes a great emphasis on Artificial Intelligence technologies implementation in civil and military areas. The aim of the article is to model relation between the Ukrainian Ministry of Defense budget programs and strategic goals and tasks. The classical budget analysis methodology is extended with NLP technics. The analysis is performed for defense budget program 2101020 - Ensuring the activities of the Armed Forces of Ukraine, training of personnel and troops, medical support of personnel, military service veterans and their family members, and war veterans. Either TF-IDF or more advanced NLP methods are used along with Python libraries and packages. It is found that some goals intersect with each other by semantic similarity. Despite the lack of data, we build proof of concept machine learning model and proved its effectiveness.

Sentiment Analysis of Oppenheimer Movie Reviews: Naïve Bayes Algorithm for Public Opinion

The development of information and communication technology has revolutionized the way people consume and engage with media, particularly in the realm of film. Online platforms such as Netflix, Amazon Prime Video, and YouTube have transformed movie consumption habits, providing a vast array of options for viewers to explore and enjoy. A crucial aspect of this digital landscape is the proliferation of movie reviews, which serve as valuable guides for users seeking to discover films aligned with their preferences. However, the abundance of reviews, often varying in quality and objectivity, necessitates tools capable of effectively processing and understanding these textual data. This research delves into sentiment classification of Oppenheimer movie reviews, utilizing the Naive Bayes algorithm to categorize reviews into positive, negative, and neutral sentiments. The dataset comprising audience reviews and numerical ratings undergoes preprocessing using the TF-IDF method to facilitate numerical representation. Subsequently, the Naïve Bayes algorithm is trained on this processed data to accurately classify sentiments. The model demonstrates exceptional performance, achieving an accuracy rate of 97.45% in distinguishing between positive, negative, and neutral sentiments within Oppenheimer movie reviews. This study underscores the efficacy of the Naive Bayes algorithm in sentiment classification and emphasizes the significance of employing techniques like TF-IDF for enhancing sentiment analysis in the domain of movie reviews.

Improving fake news detection on social networks through a novel machine learning approach

Recent years have seen an increase in societal information, such as commercial advertisements, celebrity information, political news, etc., due to the growth of online social networks. Some social media sites including Facebook, Instagram, and Twitter, were impacted by Fake News (FN) spread by their users. Nevertheless, some individuals propagate false information using words, images, and videos to increase their connections and reputation. The most current information is unreliable and misleads other members on an online social network. Because FN is circulated to encourage users to trust disinformation, it’s hard for detection algorithms to recognize false information according to uploaded material. In this research, we will provide an autonomous false news detection strategy called Grey Wolf optimized Adaptive Bidirectional Gated Recurrent Unit (GWO+ABi-GRU) that can identify FN on Twitter. The TF-IDF method translates textual material into a mathematical representation that accurately captures the value of terms and their context in the dataset. We improve the overall efficiency of the BiGRU by applying this GWO, which leads to increased accuracy when spotting false news. The experimental examination of real-world data shows that the proposed FN detection strategy GWO+ABi-GRU outperforms traditional approaches regarding recall, accuracy, F1 score, and precision. This enhanced performance proves essential in preventing the circulation of false news on social media, protecting the integrity of information sources, and mitigating the potential damage caused by false information.

Deep learning approach to detect cyberbullying on twitter

AbstractIn recent years, especially children and adolescents have shown increased interest in social media, making them a potential risk group for cyberbullying. Cyberbullying posts spread very quickly, often taking a long time to be deleted and sometimes remaining online indefinitely. Cyberbullying can have severe mental, psychological, and emotional effects on children and adolescents, and in extreme cases, it can lead to suicide. Turkey is among the top 10 countries with the highest number of children who are victims of cyberbullying. However, there are very few studies conducted in the Turkish language on this topic. This study aims to identify cyberbullying in Turkish Twitter posts. The Multi-Layer Detection (MLP) based model was evaluated using a dataset of 5000 tweets. The model was trained using both social media features and textual features extracted from the dataset. Textual features were obtained using various feature extraction methods such as Bag of Words (BOW), Term Frequency-Inverse Term Frequency (TF-IDF), Hashing Vectorizer, N-gram, and word embedding. These features were utilized in training the model, and their effectiveness was evaluated. The experiments revealed that the features obtained from TF-IDF and unigram methods significantly improved the model’s performance. Subsequently, unnecessary features were eliminated using the Chi-Square feature selection method. The proposed model achieved a higher accuracy of 93.2% compared to machine learning (ML) methods used in previous studies on the same dataset. Additionally, the proposed model was compared with popular deep learning models in the literature, such as LSTM, BLSTM, and CNN, demonstrating promising results.

Automating the Extraction of Words and Topics in Indonesian Using the Term Frequency-Inverse Document Frequency Algorithm and Latent Dirichlet Allocation

Keyword extraction and topic modeling in the analysis of Gojek user reviews in Indonesian are very important. By understanding user preferences and needs through keyword extraction, as well as grouping user reviews into different topics through topic modeling, stakeholders can use the information to further improve services. This research uses TF-IDF and LDA approaches to analyze text data from Gojek user reviews and feedback. The data spans from Nov 5, 2021, to Jan 2, 2024, totaling 225,002 rows. Each row includes username, content, time, and app version. The focus is on content reviews. The average length is 8 words, with a maximum of 104 and a minimum of a few words. The variability indicates a non-normal distribution. Preprocessing is conducted to maintain topic analysis accuracy. The TF-IDF method is used to extract relevant keywords, while the LDA approach is used to model the topics in user reviews. The topic analysis reveals patterns in Gojek user reviews. The first topic discusses experience, services, and affordable pricing. The second emphasizes app usability and benefits. The third relates to promos, discounts, and vouchers. The fourth reflects positive evaluations of service quality. However, the fifth topic highlights high costs and app issues. The sixth underscores overall user satisfaction and service convenience. Testing on the topic model yielded a coherence level of 0.509, indicating that the model's topics demonstrate a good level of consistency in finding relevant topics from Gojek user review data. The use of a combination of TF-IDF and LDA in Indonesian text analysis, particularly in the context of Gojek user reviews, is an important step in enhancing understanding and utilization of text data to improve overall user experience.

Malware Detection for Internet of Things Using One-Class Classification.

The increasing usage of interconnected devices within the Internet of Things (IoT) and Industrial IoT (IIoT) has significantly enhanced efficiency and utility in both personal and industrial settings but also heightened cybersecurity vulnerabilities, particularly through IoT malware. This paper explores the use of one-class classification, a method of unsupervised learning, which is especially suitable for unlabeled data, dynamic environments, and malware detection, which is a form of anomaly detection. We introduce the TF-IDF method for transforming nominal features into numerical formats that avoid information loss and manage dimensionality effectively, which is crucial for enhancing pattern recognition when combined with n-grams. Furthermore, we compare the performance of multi-class vs. one-class classification models, including Isolation Forest and deep autoencoder, that are trained with both benign and malicious NetFlow samples vs. trained exclusively on benign NetFlow samples. We achieve 100% recall with precision rates above 80% and 90% across various test datasets using one-class classification. These models show the adaptability of unsupervised learning, especially one-class classification, to the evolving malware threats in the IoT domain, offering insights into enhancing IoT security frameworks and suggesting directions for future research in this critical area.

Detection of COVID-19 Anti-Vaccination from Twitter Data Using Deep Learning and Feature Selection Approaches

The COVID-19 pandemic has evolved into a crisis significantly impacting health, the economy, and social life worldwide. During this crisis, anti-vaccination sentiment poses a considerable obstacle to controlling the epidemic and the effectiveness of vaccination campaigns. This study aimed to detect COVID-19 anti-vaccination sentiment from Twitter data using a combination of deep learning and feature selection approaches. The proposed method integrates a deep learning model with feature selection techniques to identify anti-vaccination sentiment by pinpointing important features in text data. Hybrid TF-IDF and N-gram methods were utilized for feature extraction, followed by Chi-square feature selection. The dataset comprises Twitter text data and two labels. The Synthetic Minority Oversampling Technique (SMOTE) was applied to balance the labels. Long Short-Term Memory (LSTM), a deep learning architecture, was employed for the classification process. The experimental results, obtained by leveraging the proposed feature extraction, feature selection, and LSTM methods, achieved the highest accuracy value of 99.23%. These findings demonstrate the proposed methods' success in effectively detecting COVID-19 anti-vaccination sentiment in text data. The study's results can offer valuable insights for developing health policies and public information strategies, presenting a new and powerful tool for detecting anti-vaccine sentiment in planning vaccination campaigns and public health interventions.

Medical Record Document Search with TF-IDF and Vector Space Model (VSM)

The growth of medical record documents is increasing over time, and the various types of diseases and therapies needed are increasing. However, this has not been followed by an effective and efficient search process. This study aims to deal with search problems that often take a long time with search results that are not necessarily as expected by building a search model for medical record documents using the vector space model (VSM) and TF-IDF methods. The VSM method allows retrieval of results that are not the same as the search queries entered by the user but are expected to provide still results relevant to the user's desired needs. The model development process was taken based on the data in the FS_ANAMNESA and FS_DIAGNOSA columns, followed by preprocessing, which consists of deleting blank lines, lowercase, removing punctuation marks, HTML tags, stop words, excess spaces between words, and normalizing typo words, then forming a TF-IDF matrix based on the frequency of occurrence of each word feature, and followed by the calculation of the similarity value of the search query compared to medical record documents based on the cosine similarity formula. The retrieval results were all columns of each existing medical record document and were sorted based on 10 rows with the highest similarity value. The model evaluation results were based on 1000 medical record documents and tested with 20 search queries in this study, which gave an average precision value of 0.548 and an average recall value of 0.796.

Content-based Filtering Movie Recommender System Using Semantic Approach with Recurrent Neural Network Classification and SGD

The application of recommendation systems has been applied in various types of platforms, especially applications for watching movies such as Netflix and Disney+. The recommendation system is purposed to make it easier for users, especially in choosing a movie because currently the number of movie productions is increasing every day. This research proposed a CBF movie recommendation system by comparing the performance of several semantic methods to be able to get the best rating prediction results. In order to improve the performance quality to get the best rating prediction results, this research utilized semantic feature methods by comparing the performance of the evaluation results produced by the TF-IDF method and word embedding applications, such as BERT, GPT-2, RoBERTa, and implemented RNN model to classify the results of rating prediction. The data were used to generate the recommendation system by involving 854 data movie and 39 accounts with a total of 34,056 movie reviews on Twitter. This research has succeeded in getting a method that produced rating predictions, namely RoBERTa. In the classification process with the RNN model and SGD optimization, the measurement results with confusion matrix by classifying the RoBERTA rating prediction obtained an evaluation value of 0.6514 loss, 95.59% accuracy, and 0.6514 precision.

AI powered Cyberbullying Detection Model

Cyberbullying has arisen as an unavoidable and concerning issue via virtual entertainment stages, influencing the psychological well-being and prosperity of people around the world. To resolve this issue, this study proposes a cyberbullying recognition framework utilizing the K-SVM calculation. Utilizing the force of AI, the framework means to consequently distinguish and signal occurrences of cyberbullying progressively web-based entertainment content. The improvement of the location framework starts with the assortment and naming of a thorough dataset containing instances of cyberbullying and non-cyberbullying posts or remarks. After pre-handling the text information by eliminating unessential data, changing message over completely to lowercase, and tokenizing it, significant highlights are removed utilizing the pack of-words or TF-IDF methods. These changed element vectors act as contributions for preparing the K-SVM classifier, which tries to find the ideal hyper plane for successfully recognizing cyberbullying from non-cyberbullying content. The exhibition of the K-SVM model is assessed utilizing a different testing dataset, with measurements, for example, exactness, accuracy, review, F1-score, and ROC-AUC broke down to survey its viability in distinguishing cyberbullying cases. Model calibrating is led through trial and error with different K-SVM hyper boundaries and cross-approval methods to upgrade the framework's exhibition. Keywords: Cyberbullying, social media, Online harassment

Impact of oxymoron features and deep learning techniques in the detection of sarcastic contents

Even though various features have been investigated in the detection of figurative language, oxymoron features have not been considered in the classification of sarcastic content. The main objective of this work is to present a system that can automatically classify sarcastic phrases in multi-domain data. This multi-domain dataset consisting of 67850 sarcastic and non-sarcastic data is collected from various websites to identify sarcastic or non-sarcastic utterances. Multiple approaches are examined in this work to improve sarcasm identification: 1. A Combination of fasttext embedding, syntactic, semantic, lexical n-gram, and oxymoron features 2. TF-IDF feature weighting scheme 3. Three machine learning algorithms (SVM, Multinomial Naïve Bayes, and Random Forest), three deep learning algorithms (CNN, LSTM, MLP), and one ensemble model (CNN + LSTM) The CNN + LSTM model achieves a Precision of 91.32%, Recall of 92.85%, F-Score of 92.08%, accuracy of 92.01%, and Kappa of 0.84 by combining the fasttext embedding, bigram, syntactic, semantic, and oxymoron features with TF-IDF method. These experimental results show CNN + LSTM with a combination of all features outperforms the other algorithms in classifying the sarcasm in both datasets. The sarcasm classification performance of our dataset and another sarcasm news dataset was compared while applying the above model.

Optimization of traditional methods for determining the similarity of project names and purchases using large language models

The subject of the study is the analysis and improvement of methods for determining the relevance of project names to the information content of purchases using large language models. The object of the study is a database containing the names of projects and purchases in the field of electric power industry, collected from open sources. The author examines in detail such aspects of the topic as the use of TF-IDF and cosine similarity metrics for primary data filtering, and also describes in detail the integration and evaluation of the effectiveness of large language models such as GigaChat, GPT-3.5, and GPT-4 in text data matching tasks. Special attention is paid to the methods of clarifying the similarity of names based on reflection introduced into the prompta of large language models, which makes it possible to increase the accuracy of data comparison. The study uses TF-IDF and cosine similarity methods for primary data analysis, as well as large GigaChat, GPT-3.5 and GPT-4 language models for detailed verification of the relevance of project names and purchases, including reflection in model prompta to improve the accuracy of results. The novelty of the research lies in the development of a combined approach to determining the relevance of project names and purchases, combining traditional methods of processing text information (TF-IDF, cosine similarity) with the capabilities of large language models. A special contribution of the author to the research of the topic is the proposed methodology for improving the accuracy of data comparison by clarifying the results of primary selection using GPT-3.5 and GPT-4 models with optimized prompta, including reflection. The main conclusions of the study are confirmation of the prospects of using the developed approach in the tasks of information support for procurement processes and project implementation, as well as the possibility of using the results obtained for the development of text data mining systems in various sectors of the economy. The study showed that the use of language models makes it possible to improve the value of the F2 measure to 0.65, which indicates a significant improvement in the quality of data comparison compared with basic methods.

Analyzing Amazon Products Sentiment: A Comparative Study of Machine and Deep Learning, and Transformer-Based Techniques

In recent years, online shopping has surged in popularity, with customer reviews becoming a crucial aspect of the decision-making process. Reviews not only help potential customers make informed choices, but also provide businesses with valuable feedback and build trust. In this study, we conducted a thorough analysis of the Amazon reviews dataset, which includes several product categories. Our primary objective was to accurately classify sentiments using natural language processing, machine learning, ensemble learning, and deep learning techniques. Our research workflow encompassed several crucial steps. We explore data collection procedures; preprocessing steps, including normalization and tokenization; and feature extraction, utilizing the Bag-of-Words and TF–IDF methods. We conducted experiments employing a variety of machine learning algorithms, including Multinomial Naive Bayes, Random Forest, Decision Tree, and Logistic Regression. Additionally, we harnessed Bagging as an ensemble learning technique. Furthermore, we explored deep learning-based algorithms, such as CNNs, Bidirectional LSTM, and transformer-based models, like XLNet and BERT. Our comprehensive evaluations, utilizing metrics such as accuracy, precision, recall, and F1 score, revealed that the BERT algorithm outperformed others, achieving an impressive accuracy rate of 89%. This research provides valuable insights into the sentiment analysis of Amazon reviews, aiding both consumers and businesses in making informed decisions and enhancing product and service quality.

Indonesian Hate Speech Text Classification Using Improved K-Nearest Neighbor with TF-IDF-ICSρF

Purpose: Freedom in social media gives rise to the possibility of disturbing users through the sentences they send, which is limited by the Electronic Information and Transactions Law (UU ITE). This research aims to find an effective method for classifying hate speech text data, especially in Indonesian, with many categories expected to minimize this case.Methods: This study used 1.000 data from Twitter with five labels, including religion, race, physical, gender and other (invective or slander). The process started with several steps of preprocessing, data transformation using TF-IDF-ICSρF term weighting and data mining using an Improved KNN algorithm. Then, the results were compared with the TF-IDF and KNN methods to evaluate the differences.Result: Using TF-IDF-ICSρF and Improved KNN algorithms gets an average accuracy value of 88.11%, 17.81% higher compared with the same data and parameters to the K-Nearest Neighbor and TF-IDF algorithms, which get results of 70.30%.Novelty: Based on the comparison results, TF-IDF-ICSρF and Improved KNN methods can effectively classify hate speech sentences that have many labels with fairly good accuracy.

Sentiment Analysis of Students’ Feedback on Faculty Online Teaching Performance Using Machine Learning Techniques

The pandemic has given rise to challenges across different sectors, particularly in educational institutions. The mode of instruction has shifted from in-person to flexible learning, leading to increased stress and concerns for key stakeholders such as teachers, parents, and students. The ongoing spread of diseases has made in-person classes unfeasible. Even if limited face to face classes will be allowed, online teaching is deemed to remain a practice to support instructional delivery to students. Therefore, it is essential to understand the challenges and issues encountered in online teaching, particularly from the perspective of students. This knowledge is crucial for supervisors and administrators, as it provides insights to aid in planning intervention measures. These interventions can support teachers in enhancing their online teaching performance for the benefit of their students. A process that can be applied to achieve this goal is sentiment analysis. In the field of education, one of the applications of sentiment analysis is in the evaluation of faculty teaching performance. It has been a practice in educational institutions to periodically assess their teachers’ performance. However, it has not been easy to take into account the students’ comments due to the lack of methods for automated text analytics. In line with this, techniques in sentiment analysis are presented in this study. Base models such as Naïve Bayes, Support Vector Machines, Logistic Regression, and Random Forest were explored in experiments and compared to a combination of the four called ensemble. Outcomes indicate that the ensemble of the four outperformed the base models. The utilization of Ngram vectorization in conjunction with ensemble techniques resulted in the highest F1 score compared to Count and TF-IDF methods. Additionally, this approach achieved the highest Cohen’s Kappa and Matthews Correlation Coefficient (MCC), along with the lowest Cross-entropy, signifying its preference as the model of choice for sentiment classification. When applied in conjunction with an ensemble, Count vectorization yielded the highest Cohen’s Kappa and Matthews Correlation Coefficient (MCC) and the lowest Cross-entropy loss in topic classification. Visualization techniques revealed that 65.4% of student responses were positively classified, while 25.5% were negatively classified. Meanwhile, predictions indicated that 47% of student responses were related to instructional design/delivery, 45.3% described the personality/behavior of teachers, 3.4% focused on the use of technology, 2.9% on content, and 1.5% on student assessment.