Research Publications
Permanent URI for this communityhttps://rda.sliit.lk/handle/123456789/4194
This main community comprises five sub-communities, each representing the academic contribution made by SLIIT-affiliated personnel.
Browse
13 results
Filters
Advanced Search
Filter by
Settings
Search Results
Publication Open Access QPred: A Lightweight Deep Learning-Based Web Pipeline for Accessible and Scalable Streamflow Forecasting(Tech Science Press, 2026) Makumbura, R.K; Wijesundara, H; Sajindra, H; Rathnayake, U; Kumar, V; Duraibabu, D; Sen, SAccurate streamflow prediction is essential for flood warning, reservoir operation, irrigation scheduling, hydropower planning, and sustainable water management, yet remains challenging due to the complexity of hydrological processes. Although data-driven models often outperform conventional physics-based hydrological modelling approaches, their real-world deployment is limited by cost, infrastructure demands, and the interdisciplinary expertise required. To bridge this gap, this study developed QPred, a regional, lightweight, cost-effective, web-delivered application for daily streamflow forecasting. The study executed an end-to-end workflow, from field data acquisition to accessible web-based deployment for on-demand forecasting. High-resolution rainfall data were recorded with tipping-bucket gauges and loggers, while river water depth in the Aglar and Paligaad watersheds was converted to discharge using site-specific rating curves, resulting in a daily dataset of precipitation, river water level and discharge. Four DL architectures were trained, including vanilla Long Short-Term Memory (LSTM), stacked LSTM, bidirectional LSTM, and Gated Recurrent Unit (GRU), and evaluated using Nash-Sutcliffe Efficiency (NSE), Coefficient of Determination (R2), Root-Mean-Square-Error-Standard-Deviation Ratio (RSR), and Percentage Bias (PBIAS) metrics. Performance was watershed-specific, as the vanilla LSTM demonstrated the best generalisation for the Aglar watershed (R2 = 0.88, NSE = 0.82, RMSE = 0.12 during validation), while the GRU achieved the highest validation accuracy in Paligaad (R2 = 0.88, NSE = 0.88, RMSE = 0.49). All models achieved satisfactory to excellent performance during calibration (R2 > 0.91, NSE > 0.91 for both watersheds), demonstrating strong capability to capture streamflow dynamics. The highest performing models were selected and embedded into the QPred application. QPred was developed as a lightweight web pipeline, utilising Google Colab as the primary execution environment, Flask as the backend inference framework, Google Drive for artefact storage, and Ngrok for secure HTTPS tunnelling. A user-friendly front end utilises range sliders (bounded by observed minima and maxima) to gather inputs and provides discharge data along with metadata, thereby enhancing transparency. This work demonstrates that accurate, context-aware deep learning models can be delivered through low-cost, web-based platforms, providing a reproducible and scalable pipeline for hydrological applications in other watersheds and for practitioners. CopyrightPublication Open Access Enhancing Organizational Threat Profiling by Employing Deep Learning with Physical Security Systems and Human Behavior Analysis(Science and Information Organization, 2025) Senevirathna D.H; Gunasekara W.M.M; Gunawardhana K.P.A.T; Ashra M.F.F; Fernando, H; Abeywardena, K. YOrganizations need a comprehensive threat profiling system that uses cybersecurity methods together with physical security methods because advanced cyber-threats have become more complex. The objective of this study is to implement deep learning models to boost organizational threat identification via human behavior assessment and continuous surveillance activities. Our method for human behavior analysis detects insider threats through assessments of user activities that include logon patterns along with device interactions and measurement of psychometric traits. CNN, together with Random Forest classifiers, has been utilized to identify behavioral patterns that indicate security threats from inside the organization. Our model uses labeled datasets of abnormal user behavior to properly differentiate between normal and dangerous user activities with high accuracy. The physical security component improves surveillance abilities through the use of MobileNetV2 for real-time anomaly detection in CCTV video data. The system receives training to detect security breaches and violent and unauthorized entry attempts, and specific security-related incidents. The combination of transfer learning and fine-tuning methodologies enables MobileNetV2 to deliver outstanding security anomaly detection alongside low power requirements, thus it fits into Security Operations Centers operations. Experiments using our framework operate on existing benchmark collection sets that assess cybersecurity, together with physical security threats. Experimental testing establishes high precision levels for detecting insider threats along with physical security violations by surpassing conventional rule-based methods. Security Operation Centers gain an effective modern threat profiling solution through the application of deep learning models. The investigation generates better organization defenses against cyber-physical threats using behavioral analytics together with intelligent surveillance systems.Publication Open Access Eco-friendly mix design of slag-ash-based geopolymer concrete using explainable deep learning(Elsevier, 2024-09) Ranasinghe, R.S.S.; Kulasooriya, W.K.V.J.B.; Perera, U S; Ekanayake, I.U.; Meddage, D.P.P.; Mohotti, D; Rathanayake, UGeopolymer concrete is a sustainable and eco-friendly substitute for traditional OPC (Ordinary Portland Cement) based concrete, as it reduces greenhouse gas emissions. With various supplementary cementitious materials, the compressive strength of geopolymer concrete should be accurately predicted. Recent studies have applied deep learning techniques to predict the compressive strength of geopolymer concrete yet its hidden decision-making criteria diminish the end-users’ trust in predictions. To bridge this gap, the authors first developed three deep learning models: an artificial neural network (ANN), a deep neural network (DNN), and a 1D convolution neural network (CNN) to predict the compressive strength of slag ash-based geopolymer concrete. The performance indices for accuracy revealed that the DNN model outperforms the other two models. Subsequently, Shapley additive explanations (SHAP) were used to explain the best-performed deep learning model, DNN, and its compressive strength predictions. SHAP exhibited how the importance of each feature and its relationship contributes to the compressive strength prediction of the DNN model. Finally, the authors developed a novel DNN-based open-source software interface to predict the mix design proportions for a given target compressive strength (using inverse modeling technique) for slag ash-based geopolymer concrete. Additionally, the software calculates the Global Warming Potential (kg CO2 equivalent) for each mix design to select the mix designs with low greenhouse emissions.Publication Embargo Image Processing-Based Solution to Repel Crop-Damaging Wild Animals(Springer, 2023-02-03) Fernando, W. P. S.; Madhubhashana, I. K.; Gunasekara, D. N. B. A.; Gogerly, Y. D.; Karunasena, A; Supunya, RTwo-thirds of Sri Lanka’s population is directly dependent on agriculture, which generates one-third of the nation’s GDP. However, crop efficiency in Sri Lanka has declined over the years due to several issues including sub-farm maintenance, destruction caused by wild animals, and unethical farming practices. Among them, the destruction caused by wild animals has led to conflicts between animals and humans causing loss of both animals and human lives in the past. There are a number of technical solutions proposed to solve the above problem, especially in the form of animal repellants. However, such solutions have several limitations, such as the small number of animal groups to be identified and the short distances they can be detected, and the lack of understanding of harmful animal populations. This research proposes an animal-repellent methodology considering several features of animals such as colors, coats, shape, and noise made by animals both in daytime and nighttime. The number of animals approaching crops is also detected and the behavior of animals is monitored to avoid false alarms. The research uses a wide range of techniques such as image processing and deep learning for the above purpose on audio, visual, and image data sets collected from the mentioned animal groups. The solution demonstrated a 90% accuracy for animal identification during the day, and 84% accuracy for animal 2 W. P. S. Fernando et al. identification at night, whereas the accuracy of studying animal behavior patterns is 90% and animal sounds were identified with 87% accuracyPublication Embargo Uncovering stress fields and defects distributions in graphene using deep neural networks(Springer, Cham, 2023-05-19) Dewapriya, M. A. N.; Rajapakse, R. K. N. D.; Dias, W. P. S.Deep learning provides a new route for developing computationally efficient predictive models for some complex engineering problems by eliminating the need for establishing exact governing equations. In this work, we used conditional generative adversarial networks (cGANs) to identify defects in graphene samples and to predict the complex stress fields created by two interacting defective regions in graphene. The required data for developing deep learning models was obtained from molecular dynamics simulations, where the numerical results of the simulations were transformed into image-based data. Our results demonstrate that the neural nets can accurately predict some complex features of the interacting stress fields. Subsequently, we used cGANs to predict defect distributions; this revealed that a cGAN could predict the existence of a crack even though it had never seen a cracked sample during the training stage. This observation clearly demonstrates the remarkable generalizability of cGANs beyond the training samples, suggesting that deep learning can be a powerful tool for solving advanced nanoengineering problems.Publication Embargo Development of an Elephant Detection and Repellent System based on Efficient Det-Lite Models(IEEE, 2023-04-03) Pemasinghe, S; Abeygunawardhana, P.K.WHuman-elephant conflict (HEC) has become a major concern in Sri Lanka that results in many unfortunate human and elephant deaths. Methods that are currently in place to mitigate HEC, such as electrical fences have undesirable consequences resulting in both human and elephant casualties. In this paper, we have proposed a method based on computer vision and deep learning that has a promising potential for detecting and repelling elephants without endangering the lives of elephants or humans. We have used EfficientDet-Lite models that provide a good compromise between accuracy and performance in order to be usable with a resource-constrained device like a Raspberry Pi.Publication Embargo E-Learning Education System For Children With Down Syndrome(Institute of Electrical and Electronics Engineers, 2022-09-16) Sampath, A.S.T; Vidanapathirana, M.W.; Gunawardana, T.B.A; Sandeepani, P.W.H.; Chandrasiri, L.H.S.S; Attanayaka, BThe World Health Organization assesses that Down Syndrome (DS) affects about 1 in 1000 births worldwide. Children with DS cannot learn, as usual, instigating numerous inadequacies that lead to formative issues such as trouble encoding information and low intelligence to interpret data for decision-making. As a superior technique for these kids' intercom-municating and logical intellect, free-hand sketch drawing, Voice training, and word prediction activities can be success-fully utilized. As the best way to express the mindset of such chil-dren, introducing an E-Learning system makes a friendlier ac-tivity than learning about the past. Because of the improvement of Artificial intelligence and its encouragement, E-Learning-re-lated exploration and applications are moving at an enormous advancement rate. The main objective of this project is to de-velop a reliable and efficient approach to predicting the devel-opment of DS children. Classifying and identifying those hand-written images and voice samples and those samples are given by children with DS compared to the teacher through the construction of a model structure. This research project specially considered local down syndrome children's hand-drawn images, voice samples, letters, numbers, and words as the input. As a result, it gives accuracy and similarity with the teacher's sam-ples and relates parts in the down syndrome children's samples. The system uses artificial intelligence technologies. Through that, the knowledge capacity of the DS children and their con-veyed articulation of that knowledge can be assessed for additional correlations and investigation.Publication Embargo AI Solution to Assist Online Education Productivity via Personalizing Learning Strategies and Analyzing the Student Performance(Institute of Electrical and Electronics Engineers, 2022-10-29) Liyanage, M.L.A.P.; Hirimuthugoda, U.J; Liyanage, N.L.T.N.; Thammita, D.H.M.M.P; Koliya Harshanath Webadu Wedanage, D; Kugathasan, A; Thelijjagoda, SHigher productivity in online education can be attained by consistent student engagement and appropriate use of learning resources and methodologies in the form of audio, video, and text. Lower literacy rates, decreased popularity, and unsatisfactory end-user goals can result from unbalanced or inappropriate use of the aforementioned. Prior studies mainly focused on identifying and separating the elements affecting the quality of online education and pinpointing the students' preferred learning styles outside of in-person and online instruction. This has not been able to clearly show how to enhance and customize the online learning environment in order to benefit the aforementioned criteria. This case study will primarily concentrate on elements that can be personalized and optimized to improve the quality of online education. With the aid of various algorithms like logistic regression,Support Vector Machines (SVM), time series forecasting (ARIMA), deep neural networks, and Recurrent Neural Networks (RNN), which make use of machine learning and deep learning techniques, the ultimate result has been attained. To increase application and accuracy, the newly presented technique will then be presented as a web-based software application. Contrary to what is commonly believed, this applied research proposes a new all-in-one Learning Management System (LMS) for students and tutors that acts as a central hub of all the learning resources.Publication Embargo Deep Learning for Code-Mixed Text Mining in Social Media: A Brief Review(Springer, Cham, 2022-09-19) Panchendraraja, R; Saxena, AThe advent of social media in day-to-day life has made communications between people more often and easier than ever before. Analyzing the content in social media has opened up a massive amount of research and commercial opportunities. However, the content in social media is noisy and multi-lingual, which postures computational challenges ahead. Especially, the non-native English speakers and writers tend to mix their native language with English while generating social media content. Thus it requires a comprehensive prepossessing of text, including the identification of language for many language processing applications. In the area of language processing, deep learning has shown to be very successful, and the latest research works have witnessed the adoption of deep learning solutions to cater to the challenges in analyzing code-mixed text. Here, we highlight a comprehensive study of deep learning techniques used for analyzing the code-mix text of social media to understand the state-of-the-art and existing research challenges. We will discuss several applications of code-mixed text analysis and future directions.Publication Embargo Characterizing fracture stress of defective graphene samples using shallow and deep artificial neural networks(Pergamon, 2020-08-15) Dewapriya, M. A. N; Rajapakse, R. K. N. D; Dias, W. P. SAdvanced machine learning methods could be useful to obtain novel insights into some challenging nanomechanical problems. In this work, we employed artificial neural networks to predict the fracture stress of defective graphene samples. First, shallow neural networks were used to predict the fracture stress, which depends on the temperature, vacancy concentration, strain rate, and loading direction. A part of the data required to model the shallow networks was obtained by developing an analytical solution based on the Bailey durability criterion and the Arrhenius equation. Molecular dynamics (MD) simulations were also used to obtain some data. Sensitivity analysis was performed to explore the features learnt by the neural network, and their behaviour under extrapolation was also investigated. Subsequently, deep convolutional neural networks (CNNs) were developed to predict the fracture stress of graphene samples containing random distributions of vacancy defects. Data required to model CNNs was obtained from MD simulations. Our results reveal that the neural networks have a strong ability to predict the fracture stress of defective graphene under various processing conditions. In addition, this work highlights some advantages as well as limitations and challenges in using neural networks to solve complex problems in the domain of computational materials design.