NSPACE

Welcome to NSBM Green University e-Repository!

One Platform for NSBM Publications

Photo by @inspiredimages

Communities in DSpace

Select a community to browse its collections.

Now showing 1 - 5 of 5

Recent Submissions

Deep Hair-Net: A Deep Learning Approach for Diagnosing Scalp and Hair Diseases with Treatment Recommendations
(NSBM Green University, 2025-11-12) Malshani, B.; Indurangala, I.G.
Scalp and hair diseases have a significant impact on individuals' physical, emotional, and social well-being. This research introduces 'Deep Hair-Net,' a novel system for the automated diagnosis of common scalp ailments coupled with actionable treatment recommendations. The 'Deep Hair-Net' system employs a hybrid deep learning architecture, fusing the strengths of MobileNetV2 and Inception V3 through transfer learning. This model was trained and validated on a curated dataset of 10,132 images (sourced from DermNet, ISIC, and Kaggle) covering ten distinct scalp disease classes, including dandruff, alopecia, and psoriasis. The hybrid model demonstrated high performance, achieving 95.5% classification accuracy, an F1-score of 0.92, and an AUC-ROC of 0.96. A key feature of the system is its integrated CSV-based recommendation engine, which maps the diagnostic output to specific management suggestions. This research contributes a scalable and robust AI system that closes the gap between dermatological diagnostics and practical treatment advice, offering significant value for teledermatology and resource limited environments.
Machine Learning Approaches for Short-Term Rooftop PV Forecasting in Tropical Climates: A Systematic Review
(NSBM Green University, 2025-11-12) Gunathilaka, A.S.A.; Wijesekara, D.
In many tropical regions, rooftop photovoltaic (PV) systems are pivotal for sustainable energy transitions. Nevertheless, grid stability and operational planning are complicated due to their variable output. With a focus on tropical climates and developing-country contexts, particularly in Sri Lanka, this systematic review combines machine learning (ML), deep learning (DL), and hybrid forecasting approaches for short-term (1–24 hours) Rooftop PV prediction. Classifying based on algorithm, data requirements, system scale, and climatic context, 32 peer reviewed studies were found through a systematic search across major academic databases (2018–2025). Key findings demonstrate that while hybrid ML–DL models (e.g., CNNBiLSTM), optimization-enhanced networks, and ensemble frameworks outperform conventional statistical baselines, they encounter real-world limitations, including limited validation on distributed rooftop systems, high computational cost, and data scarcity. Promising directions include lightweight and interpretable hybrid models, tropical dataset development, and the integration of explainable AI with uncertainty quantification. Overall, this review provides a practical roadmap for researchers and practitioners aiming to design scalable, dependable PV forecasting systems in resource-constrained tropical environments.
University Human-Centered Supervision Platform for Student and Supervisor Collaboration in Research
(NSBM Green University, 2025-11-12) Paranagama, R.A.; Kankanamge, P.
Higher education institutions are realizing more the importance structured research supervision is to timely advancement, efficient communication, and high-quality results. Current supervision procedures at the National School of Business Management (NSBM) rely on unofficial channels including emails, WhatsApp groups, and occasional meetings, which frequently lead to misunderstandings, delayed feedback, and inadequate tracking of students' progress. This study suggests a human-centered, centralized supervision management system that combines a mobile application for students with a safe web dashboard for supervisors and administrators. To improve responsibility and teamwork, the platform facilitates scheduled meeting scheduling, automated deadline reminders, structured document submission, and transparent feedback sharing. A quantitative assessment of final-year students and NSBM lecturers revealed significant issues with supervision procedures, which influenced its design features, including chatbot support and notification alerts. The results show that most participants experience problems with scheduling, feedback delays, and monitoring, underscoring the necessity of a single digital solution. It is anticipated that the suggested platform will guarantee the timely completion of research projects, improve productivity, and close communication gaps. All things considered, this study advances science by showing how a supervision model based on human–computer interaction (HCI) might improve academic cooperation and simplify research management in higher education settings.
Ensemble Deep Learning Approaches for Multiclass Classification of Hip Region Fractures in X-Ray Images
(NSBM Green University, 2025-11-12) Minuja, K.; Luxshi, K.; Abishethvarman, V.; Prasanth, S.; Kumara, B.T.G.S.
Hip region fractures, including pelvic, femoral neck, intertrochanteric, and subtrochanteric fractures, are critical medical conditions, especially when diagnosed early. These fractures impair mobility, increase risks, and cause complications. Early diagnosis using X-ray imaging is vital for effective treatment. Recent advances in computer vision, particularly ensemble pretrained models, have revolutionized fracture detection by combining various models to improve classification accuracy and stability. This research developed and evaluated ensemble deep learning methods for multiclass classification of hip fractures on X-ray images. The dataset consists of 1000 X-ray images from Sri Lankan hospitals (20222023), categorized into two types: non-fracture and fracture. Preprocessing and data augmentation techniques are used to increase dataset diversity. The data was split into 70:15:15 for training, validation, and testing to evaluate performance. The pretrained model architectures include ResNet- 101, ResNet-50, EfficientNetB0, and EfficientNetV2, with ResNet-10 taken with different levels of parameterization. ResNet101 achieves the highest test accuracy of 0.8000, followed by ResNet- 50 (0.7786), EfficientNetB0 (0.7286), and EfficientNetV2 (0.7500). These pretrained models are induced as ensemble learning models and enhance multiclass hip fracture classification, yielding more accurate results compared to customized vision models. This approach has potential clinical applications, aiding early and reliable diagnosis. Further, it can extend to differentiate the components of the hip region individually with sophisticated data augmentation techniques that help for the classification. This research proves that pretrained models can be effective in biomedical rather than building and training them from scratch.
Deep Learning and Ensemble Models for Brain Tumor Classification using Medical Imaging
(NSBM Green University, 2025-11-12) Vasavan, M.; Luxshi, K.; Nadika, C.
The proposed research focuses on the early and accurate detection of brain tumors using advanced deep learning and ensemble models based on medical imaging data to enhance diagnostic outcomes. The dataset consists of 4234 MRI images collected from publicly available sources and Kaggle repositories, comprising 1592 no-tumor, 649 glioma, 999 meningioma, and 994 pituitary tumor images. Data augmentation and cleaning techniques were applied to increase diversity and reduce class imbalance. The dataset was divided into 80% for training and 20% for testing, and models such as Convolutional Neural Networks (CNN), MobileNetV2, and Multi-Layer Perceptron (MLP) were employed. To improve robustness, 5-fold cross- validation was conducted, achieving a mean testing accuracy of 0.97 ± 0.01, with precision and recall values exceeding 0.95. The ensemble model adopts a meta-model stacking architecture, where predictions from CNN and MobileNetV2 are combined as inputs to an MLP classifier, achieving significantly higher results than baseline models such as VGG16 (0.90) and ResNet50 (0.92). The system was implemented as a Streamlit-based web application, demonstrating real-time clinical usability and scalability in hospital environments. The proposed model’s high accuracy, stability, and lightweight design make it an efficient, fast, and reliable solution to support clinicians in brain tumor diagnosis and improve patient care.