Maizatul Akmar Ismail
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Maizatul Akmar Ismail has a total of 89 co-authors. Here are the top co-authors by number of shared publications:
- Khalid Haruna
Department of Computer Science, Faculty of Computer Science and Information Technology, Bayero University, Kano, Nigeria
5 shared publications - Tutut Herawan
Universiti Malaya
5 shared publications - Haruna Chiroma
College of Computer Science and Engineering, University of Hafr Batin, Hafr Batin, Saudi Arabia
4 shared publications - Mohammad Imran
Department of Information Systems, Faculty of Computer Science and Information Technology, University of Malaya, Kuala Lumpur, Malaysia
3 shared publications - Suraya Hamid
Faculty of Computer Science and Information Technology, University of Malaya, Kuala Lumpur 50603, Malaysia
3 shared publications - Amirrudin Kamsin
Faculty of Computer Science & Information Technology, Universiti Malaya, Kuala Lumpur, Malaysia
2 shared publications
Latest Publications
- 8 August 2026
- Published by Springer Nature
- p. 111-123
Enhancing Multi-Objective Eco-Efficient Flight Route Optimization Under Dynamic Constraint
- 8 August 2026
- Published by Springer Nature
- p. 111-123
Enhancing Multi-Objective Eco-Efficient Flight Route Optimization Under Dynamic Constraint
- 31 July 2026
- Published by Springer Nature in Journal of Medical Systems
- Vol. 50 (1) , 120
Automated Multi-Label Adverse drug Reaction Extraction from Malaysia’s National Pharmacovigilance Narratives
- 31 July 2026
- Published by Springer Nature in Journal of Medical Systems
- Vol. 50 (1) , 120
Automated Multi-Label Adverse drug Reaction Extraction from Malaysia’s National Pharmacovigilance Narratives
- Open Access
- 29 January 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 13, 24622-24641
Comprehensive Review of Meta-Learning Methods for Cold-Start Issue in Recommendation Systems
The cold-start issue in recommendation systems refers to the challenge of recommending items or users when minimal or no prior data is available. Meta-learning methods have emerged as a response to this challenge due to their ability to transfer prior knowledge to recommendation tasks. However, meta-learning techniques are still new, and a general literature review is missing. This paper reviews the existing literature on meta-learning techniques specifically designed to solve the cold-start issue in recommender systems. A systematic review of the literature published between 2018 and June 2024 was conducted, identifying only experimental papers that use meta-learning methods to solve the cold-start issue. Advances, strengths, and weaknesses of such methods were analyzed, and possible research directions for the future were identified. The results demonstrate the application of model-independent meta-learning (MAML) and other techniques such as optimization-based methods, few-shot learning frameworks, and gradient-based meta-learning methods to solve the cold start problem. It also shows how meta-learning improvement can be achieved by combining different strategies, using transfer learning, and effectively implementing the strategy. Some of the areas for further research are also listed. In summary, this work contributes to verifying the central mechanisms of the recently proposed meta-learning models for further research in dealing with the cold-start issue. Insights into user-item interactions, critical applications, and evaluation standards are provided.
- 29 January 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 13, 24622-24641
Open AccessComprehensive Review of Meta-Learning Methods for Cold-Start Issue in Recommendation Systems
The cold-start issue in recommendation systems refers to the challenge of recommending items or users when minimal or no prior data is available. Meta-learning methods have emerged as a response to this challenge due to their ability to transfer prior knowledge to recommendation tasks. However, meta-learning techniques are still new, and a general literature review is missing. This paper reviews the existing literature on meta-learning techniques specifically designed to solve the cold-start issue in recommender systems. A systematic review of the literature published between 2018 and June 2024 was conducted, identifying only experimental papers that use meta-learning methods to solve the cold-start issue. Advances, strengths, and weaknesses of such methods were analyzed, and possible research directions for the future were identified. The results demonstrate the application of model-independent meta-learning (MAML) and other techniques such as optimization-based methods, few-shot learning frameworks, and gradient-based meta-learning methods to solve the cold start problem. It also shows how meta-learning improvement can be achieved by combining different strategies, using transfer learning, and effectively implementing the strategy. Some of the areas for further research are also listed. In summary, this work contributes to verifying the central mechanisms of the recently proposed meta-learning models for further research in dealing with the cold-start issue. Insights into user-item interactions, critical applications, and evaluation standards are provided.
- Open Access
- 13 November 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 12, 172923-172942
Partition-Based Clustering Algorithms Applied to Mixed Data for Educational Data Mining: A Survey From 1971 to 2024
Educational Data Mining (EDM) is the application of data mining methods in the educational domain. In the EDM field, we see mixed data (i.e., text and number data types). Grouping or clustering such data is challenging because determining the similarity between mixed data is poorly defined. Existing partition clustering algorithms for handling such data are based on two approaches: conversion of data types, where all data variables are converted to a single data type, and a mixed one, where the similarity measures of different data types are merged by either using a weighted sum approach as in Gower’s distance or by using mixed dissimilarity function as used in the k-Medoids algorithm to define a singular similarity measure for mixed data. Such a datatype conversion causes information loss, and this aspect is not discussed in the existing research literature. This study systematically reviews the past fifty-three years i.e. from 1971 to 2024 of research works on partition clustering algorithms applied to mixed data in EDM. A review of 104 research articles noted that most partitional clustering algorithms have continuous or categorical variables but not mixed variables. Researchers and practitioners often cite the lack of continuous and categorical variables analysis methods. Therefore, developing machine learning algorithms that can handle mixed data inherently present in the educational domain is increasingly becoming important. In addition to comparative analysis and analysis based on several factors, research gaps are also identified and mentioned in this article, and future insights are outlined.
- 13 November 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 12, 172923-172942
Open AccessPartition-Based Clustering Algorithms Applied to Mixed Data for Educational Data Mining: A Survey From 1971 to 2024
Educational Data Mining (EDM) is the application of data mining methods in the educational domain. In the EDM field, we see mixed data (i.e., text and number data types). Grouping or clustering such data is challenging because determining the similarity between mixed data is poorly defined. Existing partition clustering algorithms for handling such data are based on two approaches: conversion of data types, where all data variables are converted to a single data type, and a mixed one, where the similarity measures of different data types are merged by either using a weighted sum approach as in Gower’s distance or by using mixed dissimilarity function as used in the k-Medoids algorithm to define a singular similarity measure for mixed data. Such a datatype conversion causes information loss, and this aspect is not discussed in the existing research literature. This study systematically reviews the past fifty-three years i.e. from 1971 to 2024 of research works on partition clustering algorithms applied to mixed data in EDM. A review of 104 research articles noted that most partitional clustering algorithms have continuous or categorical variables but not mixed variables. Researchers and practitioners often cite the lack of continuous and categorical variables analysis methods. Therefore, developing machine learning algorithms that can handle mixed data inherently present in the educational domain is increasingly becoming important. In addition to comparative analysis and analysis based on several factors, research gaps are also identified and mentioned in this article, and future insights are outlined.
- Open Access
- 11 November 2024
- Published by MDPI AG in Mathematics
- Vol. 12 (22) , 3513
Deep Learning-Driven Virtual Furniture Replacement Using GANs and Spatial Transformer Networks
This study proposes a Generative Adversarial Network (GAN)-based method for virtual furniture replacement within indoor scenes. The proposed method addresses the challenge of accurately positioning new furniture in an indoor space by combining image reconstruction with geometric matching through combining spatial transformer networks and GANs. The system leverages deep learning architectures like Mask R-CNN for executing image segmentation and generating masks, and it employs DeepLabv3+, EdgeConnect algorithms, and ST-GAN networks for carrying out virtual furniture replacement. With the proposed system, furniture shoppers can obtain a virtual shopping experience, providing an easier way to understand the aesthetic effects of furniture rearrangement without putting in effort to physically move furniture. The proposed system has practical applications in the furnishing industry and interior design practices, providing a cost-effective and efficient alternative to physical furniture replacement. The results indicate that the proposed method achieves accurate positioning of new furniture in indoor scenes with minimal distortion or displacement. The proposed system is limited to 2D front-view images of furniture and indoor scenes. Future work would involve synthesizing 3D scenes and expanding the system to replace furniture images photographed from different angles. This would enhance the efficiency and practicality of the proposed system for virtual furniture replacement in indoor scenes.
- 11 November 2024
- Published by MDPI AG in Mathematics
- Vol. 12 (22) , 3513
Open AccessDeep Learning-Driven Virtual Furniture Replacement Using GANs and Spatial Transformer Networks
This study proposes a Generative Adversarial Network (GAN)-based method for virtual furniture replacement within indoor scenes. The proposed method addresses the challenge of accurately positioning new furniture in an indoor space by combining image reconstruction with geometric matching through combining spatial transformer networks and GANs. The system leverages deep learning architectures like Mask R-CNN for executing image segmentation and generating masks, and it employs DeepLabv3+, EdgeConnect algorithms, and ST-GAN networks for carrying out virtual furniture replacement. With the proposed system, furniture shoppers can obtain a virtual shopping experience, providing an easier way to understand the aesthetic effects of furniture rearrangement without putting in effort to physically move furniture. The proposed system has practical applications in the furnishing industry and interior design practices, providing a cost-effective and efficient alternative to physical furniture replacement. The results indicate that the proposed method achieves accurate positioning of new furniture in indoor scenes with minimal distortion or displacement. The proposed system is limited to 2D front-view images of furniture and indoor scenes. Future work would involve synthesizing 3D scenes and expanding the system to replace furniture images photographed from different angles. This would enhance the efficiency and practicality of the proposed system for virtual furniture replacement in indoor scenes.
- Open Access
- 13 August 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 12, 138972-138984
Aspect Extraction in Domain Lexicon Generation: A New Frequency-Based Approach
Domain sentimental lexicon building become an attractive field in recent years. This is due to the increased number of users’ generated data through the internet besides the different sentiments of opinion words in different contexts. Domain lexicons mainly consist of opinion pairs and their associated sentiment. Any opinion pair is formed by a domain word and one of its associated opinion words. Therefore, to generate a domain lexicon from a domain corpus, domain word extraction is needed with their associated opinion words. One of the traditional approaches is frequency-based approaches. However, the ambiguity problem is a big concern of these approaches. This paper introduced a frequency-based equation that considers the context of the words for domain word extraction. The equation was tested on five Amazon reviews datasets and it proved its efficiency over other used frequency-based equations in terms of recall and precision. Therefore, more related lexicons to the domains were generated.
- 13 August 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 12, 138972-138984
Open AccessAspect Extraction in Domain Lexicon Generation: A New Frequency-Based Approach
Domain sentimental lexicon building become an attractive field in recent years. This is due to the increased number of users’ generated data through the internet besides the different sentiments of opinion words in different contexts. Domain lexicons mainly consist of opinion pairs and their associated sentiment. Any opinion pair is formed by a domain word and one of its associated opinion words. Therefore, to generate a domain lexicon from a domain corpus, domain word extraction is needed with their associated opinion words. One of the traditional approaches is frequency-based approaches. However, the ambiguity problem is a big concern of these approaches. This paper introduced a frequency-based equation that considers the context of the words for domain word extraction. The equation was tested on five Amazon reviews datasets and it proved its efficiency over other used frequency-based equations in terms of recall and precision. Therefore, more related lexicons to the domains were generated.
- Open Access
- 13 June 2024
- Published by Multimedia University in Journal of Informatics and Web Engineering
- Vol. 3 (2) , 199-211
Performance Evaluation of YOLO Models in Plant Disease Detection
Plant diseases significantly impact global agriculture, leading to substantial production losses and economic consequences. Timely disease detection can enhance crop yield, optimize resource utilization, reduce costs, and mitigate environmental effects, ultimately ensuring high-quality food production. Deep learning, specifically computer vision-based techniques, have proven invaluable in tasks like image classification, segmentation, and object detection. Deep Learning techniques such as You Only Look Once (YOLO) models are state of the art neural network algorithms used for accurate object detection. In this study, YOLOv5, YOLOv7 and YOLOv8 models were trained on CCL’20 dataset for citrus disease detection. Data augmentation techniques such as image translation, image scaling, flip, mosaic augmentations were implemented to improve the models’ performance during training phase. The model performance was evaluated using metric such as Mean Average Precision at 50% to 95% Intersection over Union score i.e. mAP@50-95. The results show that YOLOv8 model performs better than other variants and offers significant improvements over the benchmark performance from previous studies. The final hyper-parameter tuned model achieved 96.1% mAP@50-95 on testing data for citrus disease detection and mAP@50-95 of 95.3%, 96.0% and 97.0% for detection of Anthracnose, Melanose and Bacterial Brown Spot diseases, respectively. The trained model was able to detect single and multiple instances of same or different disease in an image showing the potential of recent YOLO models. The trained YOLOv8 model is deployed on Roboflow platform.
- 13 June 2024
- Published by Multimedia University in Journal of Informatics and Web Engineering
- Vol. 3 (2) , 199-211
Open AccessPerformance Evaluation of YOLO Models in Plant Disease Detection
Plant diseases significantly impact global agriculture, leading to substantial production losses and economic consequences. Timely disease detection can enhance crop yield, optimize resource utilization, reduce costs, and mitigate environmental effects, ultimately ensuring high-quality food production. Deep learning, specifically computer vision-based techniques, have proven invaluable in tasks like image classification, segmentation, and object detection. Deep Learning techniques such as You Only Look Once (YOLO) models are state of the art neural network algorithms used for accurate object detection. In this study, YOLOv5, YOLOv7 and YOLOv8 models were trained on CCL’20 dataset for citrus disease detection. Data augmentation techniques such as image translation, image scaling, flip, mosaic augmentations were implemented to improve the models’ performance during training phase. The model performance was evaluated using metric such as Mean Average Precision at 50% to 95% Intersection over Union score i.e. mAP@50-95. The results show that YOLOv8 model performs better than other variants and offers significant improvements over the benchmark performance from previous studies. The final hyper-parameter tuned model achieved 96.1% mAP@50-95 on testing data for citrus disease detection and mAP@50-95 of 95.3%, 96.0% and 97.0% for detection of Anthracnose, Melanose and Bacterial Brown Spot diseases, respectively. The trained model was able to detect single and multiple instances of same or different disease in an image showing the potential of recent YOLO models. The trained YOLOv8 model is deployed on Roboflow platform.
- Open Access
- 30 April 2024
- Published by Univ. of Malaya in Malaysian Journal of Computer Science
- Vol. 37 (2) , 154-168
PERFORMANCE EVALUATION OF MULTILABEL EMOTION CLASSIFICATION USING DATA AUGMENTATION TECHNIQUES
One of the challenges of emotion classification is the existence of low annotated datasets, that makes the task more complex. Certain existing datasets often suffer from imbalanced data for the emotion classes. Several data augmentation approaches can help to overcome the challenges regarding imbalanced datasets. However, the existing data augmentation techniques in emotion classification lack consideration for the contextual nuances of emotions and this area is still relatively underexplored. In this work, we study the impact of data augmentation on classification performance of three machine learning models including Logistic Regression, BiLSTM and BERT and compare frequently used methods to address the issue. Specifically, we assessed Easy Data Augmentation (EDA) and contextual Embedding-based data augmentation (BERT) on two datasets. Based on the experimental results, we combined two BERT-based augmentation techniques including insert and substitute, to generate data for minority emotion classes. Furthermore, we proposed a data augmentation method using ChatGPT. Compared to the baseline models, incorporating the BERT augmentation techniques with BERT model resulted in improvements of +4.34% and +5.56% in Macro F1 score on the SemEval-2018 and GoEmotions datasets, respectively. Moreover, the proposed augmentation technique utilizing ChatGPT yielded improvements of +3.55% and +4.83% on the same datasets.
- 30 April 2024
- Published by Univ. of Malaya in Malaysian Journal of Computer Science
- Vol. 37 (2) , 154-168
Open AccessPERFORMANCE EVALUATION OF MULTILABEL EMOTION CLASSIFICATION USING DATA AUGMENTATION TECHNIQUES
One of the challenges of emotion classification is the existence of low annotated datasets, that makes the task more complex. Certain existing datasets often suffer from imbalanced data for the emotion classes. Several data augmentation approaches can help to overcome the challenges regarding imbalanced datasets. However, the existing data augmentation techniques in emotion classification lack consideration for the contextual nuances of emotions and this area is still relatively underexplored. In this work, we study the impact of data augmentation on classification performance of three machine learning models including Logistic Regression, BiLSTM and BERT and compare frequently used methods to address the issue. Specifically, we assessed Easy Data Augmentation (EDA) and contextual Embedding-based data augmentation (BERT) on two datasets. Based on the experimental results, we combined two BERT-based augmentation techniques including insert and substitute, to generate data for minority emotion classes. Furthermore, we proposed a data augmentation method using ChatGPT. Compared to the baseline models, incorporating the BERT augmentation techniques with BERT model resulted in improvements of +4.34% and +5.56% in Macro F1 score on the SemEval-2018 and GoEmotions datasets, respectively. Moreover, the proposed augmentation technique utilizing ChatGPT yielded improvements of +3.55% and +4.83% on the same datasets.
- Open Access
- 19 March 2024
- Published by Qeios Ltd
Review of: "Perceptions of Academic Dishonesty: Insights from the University of Tehran"
The paper put forward an intriguing issue, particularly given the tumultuous impact of AI-powered writing and editing platforms/tools. It could be further improved by considering the following points : * The justification to focus on undergraduate...
- 19 March 2024
- Published by Qeios Ltd
Open AccessReview of: "Perceptions of Academic Dishonesty: Insights from the University of Tehran"
The paper put forward an intriguing issue, particularly given the tumultuous impact of AI-powered writing and editing platforms/tools. It could be further improved by considering the following points : * The justification to focus on undergraduate...
- Open Access
- 18 August 2023
- Published by MDPI AG in Sustainability
- Vol. 15 (16) , 12539
Hybrid Feature Extraction for Multi-Label Emotion Classification in English Text Messages
Emotions are vital for identifying an individual’s attitude and mental condition. Detecting and classifying emotions in Natural Language Processing applications can improve Human–Computer Interaction systems, leading to effective decision making in organizations. Several studies on emotion classification have employed word embedding as a feature extraction method, but they do not consider the sentiment polarity of words. Moreover, relying exclusively on deep learning models to extract linguistic features may result in misclassifications due to the small training dataset. In this paper, we present a hybrid feature extraction model using human-engineered features combined with deep learning based features for emotion classification in English text. The proposed model uses data augmentation, captures contextual information, integrates knowledge from lexical resources, and employs deep learning models, including Bidirectional Long Short-Term Memory (Bi-LSTM) and Bidirectional Encoder Representation and Transformer (BERT), to address the issues mentioned above. The proposed model with hybrid features attained the highest Jaccard accuracy on two of the benchmark datasets, with 68.40% on SemEval-2018 and 53.45% on the GoEmotions dataset. The results show the significance of the proposed technique, and we can conclude that the incorporation of the hybrid features improves the performance of the baseline models.
- 18 August 2023
- Published by MDPI AG in Sustainability
- Vol. 15 (16) , 12539
Open AccessHybrid Feature Extraction for Multi-Label Emotion Classification in English Text Messages
Emotions are vital for identifying an individual’s attitude and mental condition. Detecting and classifying emotions in Natural Language Processing applications can improve Human–Computer Interaction systems, leading to effective decision making in organizations. Several studies on emotion classification have employed word embedding as a feature extraction method, but they do not consider the sentiment polarity of words. Moreover, relying exclusively on deep learning models to extract linguistic features may result in misclassifications due to the small training dataset. In this paper, we present a hybrid feature extraction model using human-engineered features combined with deep learning based features for emotion classification in English text. The proposed model uses data augmentation, captures contextual information, integrates knowledge from lexical resources, and employs deep learning models, including Bidirectional Long Short-Term Memory (Bi-LSTM) and Bidirectional Encoder Representation and Transformer (BERT), to address the issues mentioned above. The proposed model with hybrid features attained the highest Jaccard accuracy on two of the benchmark datasets, with 68.40% on SemEval-2018 and 53.45% on the GoEmotions dataset. The results show the significance of the proposed technique, and we can conclude that the incorporation of the hybrid features improves the performance of the baseline models.