IJRC https://ir.kdu.ac.lk/handle/345/7540 2026-04-08T13:42:42Z 2026-04-08T13:42:42Z Jayasekera, SP Kalansooriya, LP https://ir.kdu.ac.lk/handle/345/7552 2026-02-19T07:24:27Z 2024-01-01T00:00:00Z

A Comprehensive Review of Methods Used for Health Prediction and Monitoring Utilizing an Electronic Medical Records (EMR) System Jayasekera, SP; Kalansooriya, LP In the rapidly evolving field of healthcare, Artificial Intelligence (AI) and pattern recognition play a key role in enhancing disease diagnosis and prediction. As the patient population increases, the digitalization of medical records has become essential, therefore electronic medical records were developed. This stored Electronic Medical Records (EMR) data can be used to predict possible diseases based on the symptoms stored in the system. This study delves into the integration of AI methodologies within EMR systems, providing a comprehensive review of current techniques that have been used in health prediction and monitoring using EMR data. In this paper, different AI-driven approaches were examined and compared, including Deep Learning (DL), Machine Learning (ML), and Rule-Based Methods. This paper reveals the potential of these techniques in accurately diagnosing diseases, additionally, it discusses challenges and future directions, emphasizing the need for innovative solutions to optimize EMR systems in the context of AI and pattern recognition. Several instances where AI models, such as the application of Support Vector Machine (SVM) models, achieved predictive accuracies of 86.2% and 97.33% in different cancer types, and ML models diagnosing Diabetic Retinopathy with a 92% accuracy rate were observed. Variations in the effectiveness of these technologies across different diseases were also observed, such that a technique that has high accuracy in one disease may have lower accuracy in a different disease. This paper aims to contribute to the growing body of knowledge in AI applications in healthcare, offering insights into the development of more efficient, accurate, and predictive healthcare models.

2024-01-01T00:00:00Z Kumari, HMLS https://ir.kdu.ac.lk/handle/345/7545 2026-02-19T08:11:35Z 2024-01-01T00:00:00Z

Convolutional Neural Network-Based Facial Expression Recognition: Enhanced by Data Augmentation and Transfer Learning Kumari, HMLS Facial expression recognition has emerged as a dynamic field within computer vision and human-computer interaction, finding diverse applications such as animation, social robots, personalized banking, and more. Current studies employ transfer learning models in facial expression recognition through the application of convolutional neural networks. The proposed model combines data augmentation with fine-tunned transfer learning models to get a better FER model. A comprehensive collection of training images is crucial as input to effectively train a convolutional neural network (CNN) for accurate facial expression recognition. Hence, the presented research employed data augmentation to enhance the quantity of input images derived from a pre-existing dataset. Manually employing CNN is outdated. Therefore, fine-tuned transfer learning models are used in the proposed study. Activating the final 8 layers of the transfer learning model by freezing the whole transfer learning model is the novel methodology of the proposed model. Then we vary the values of dense layers and dropout layers of the activated 8 layers, which results the fine-tuning of the transfer learning model. The CK+, The facial recognition dataset (human) datasets are used in the proposed model. Subsequently, conduct a stratified 5fold cross-validation to assess the model's performance on previously unseen data and avoid overfitting the proposed model. The method under consideration utilized transfer learning models, namely DenseNet121, DenseNet201, DenseNet169, and InceptionV3, along with fine-tuned transfer learning models applied to augmented datasets CK+, The facial recognition dataset (human) datasets. The outcomes indicate an achievement of 99.36% accuracy for the CK+ dataset, 95.14% for the facial recognition dataset (Human).

2024-01-01T00:00:00Z Senadheera, JR Madushanka, MKP Gunathilake, HRWP https://ir.kdu.ac.lk/handle/345/7544 2026-02-19T08:12:29Z 2024-01-01T00:00:00Z

Integrated Approach for Asset Price Forecasting via Prophet Model and Optimizing Investment Strategies through Genetic Algorithms Senadheera, JR; Madushanka, MKP; Gunathilake, HRWP This research presents an in-depth exploration of a wide array of algorithms, techniques, methods and models used for forecasting asset values. Significantly, the study introduces an unprecedented approach, featuring a dedicated model for precise price forecasting and another for recommending optimized strategies. By assessing and contrasting the approaches and outcomes of asset value prediction across different fields, this paper study aims to harness the power of Artificial Intelligence (AI) in forecasting asset prices and tailoring investment strategies. Implemented system integrates the Prophet Model for precise price forecasting and employs Genetic Algorithms for investment strategy generation. Through a systematic evaluation of the system, we demonstrate its capacity to provide accurate asset price predictions, outperform traditional investment strategies and mitigate risks effectively. Empirical unit testing showcased impressive results such as gold model with a 4.76% MAPE and an R-squared value of 0.9795 and oil model with notable metrics such as a Mean Absolute Error of 6.80, and Root Mean Squared Error of 10.92. Every single user, across the board, either strongly agreed or agreed that the investment recommendations provided valuable insights and 92.4% perceiving system predictions as very accurate. It further delves into the challenges and limitations, such as the quality of data used and model interpretability, underscoring the imperative for robust, compliant and interpretable forecasting models. Additionally, the study explores future directions in the domain, advocating for the expansion of asset classes and the integration of Natural Language Processing (NLP) into the system.

2024-01-01T00:00:00Z Jayaweera, MPGK Kithulwatta, WMCJT Rathnayaka, RMKT https://ir.kdu.ac.lk/handle/345/7543 2026-02-19T08:15:58Z 2024-01-01T00:00:00Z

An Approach to Examine and Recognize Anomalies on Cloud Computing Platforms with Machine Learning Concepts Jayaweera, MPGK; Kithulwatta, WMCJT; Rathnayaka, RMKT Cloud computing is one of the most rapidly growing computing concepts in today's information technology world. It connects data and applications from various geographical locations. A large number of transactions and the hidden infrastructure in cloud computing systems have presented the research community with several challenges. Among these, maintaining cloud network security has emerged as a major challenge. It is critical to address issues in the quickly changing cloud computing market in order to guarantee that businesses can fully utilize cutting-edge technology, uphold strong security protocols, and maximize operational effectiveness. Businesses that successfully navigate these obstacles can maintain their competitiveness in a dynamic digital ecosystem by improving scalability, leveraging the flexibility provided by the cloud, and adapting to technological changes with ease. Anomaly detection (or outlier detection) is the identification of unusual or suspicious data that differs significantly from the majority of the data. Research on anomaly detection in cloud network data is crucial because it enables businesses to more rapidly and efficiently recognize potential security threats, network performance concerns, and other issues. Recently, machine learning methods have demonstrated their efficacy in anomaly detection. This research aimed to introduce a novel hybrid model for anomaly detection in cloud network data and to investigate the performance of this model in comparison to other machine learning algorithms. The research was conducted with the UNSW-NB15 anomaly dataset and employed various feature selection and pre-processing techniques to prepare the data for model training. The hybrid model was built using a combination of Random Forest and SVM algorithms and the process was evaluated using metrics such as F1-Score, Recall, Precision, and Accuracy. The result showed that the hybrid model has 94.23% accuracy and a total time of 109.92s which is the combination of the train time of 100.45s and prediction time of 9.47s. The limitations of the study include the class imbalance problem in the dataset and the lack of real-world applications for testing. The research suggests future work in the application of hybrid models in anomaly detection and cloud network security and the need for further investigation into the potential benefits of such models.

2024-01-01T00:00:00Z