IJRC https://ir.kdu.ac.lk/handle/345/5293 2026-04-08T13:32:25Z 2026-04-08T13:32:25Z Fernando, RM Ilmini, WMKS Vidanagama, DU https://ir.kdu.ac.lk/handle/345/5301 2025-09-01T08:15:33Z 2022-01-10T00:00:00Z

Prediction of Air Quality Index in Colombo Fernando, RM; Ilmini, WMKS; Vidanagama, DU Air is always considered as the main critical factor on which human survival depends on. The AQI or long firmly air quality index is the index value that illustrates qualitatively the current state of the air. The substantial AQI will further menace the living creatures’ health & the living atmosphere. Terrible air quality has been a major concern in Sri Lanka, particularly in urban cities such as Colombo and Kandy. Reliable AQI prediction will assist to decrease the health risks caused by air pollution. The goal of this study has been to find the most suitable machine learning approach for predicting accurate air quality index in Colombo based upon PM2.5 particular concentration. In this study, PM2.5 concentration in Colombo had been predicted using four correlated air pollutant concentrations such as SO2, NO2, PM2.5, & PM10. The obtained dataset was pre-processed via prediction in order to improve prediction accuracy. The gathered dataset Cross-validated as according to 80% for training & 20% for testing the prediction model. Machine learning methods such as K-Nearest Neighboring, Multiple Linear-Regression, Random Forest, and Support Vector Machines were used to train and evaluate the prediction models. In the end, we achieved 83.25% accuracy for the K-Nearest Neighboring algorithm model, 84.68% accuracy for the Support Vector Machines model, 85.17% accuracy for the Random Forest model, and 41.9% accuracy for the Multiple Regression Model. Random Forest was recognized as the best appropriate prediction model after evaluating the models, with over 85% greater accuracy.

2022-01-10T00:00:00Z Goonathilake, MDPP Kumara, PPNV https://ir.kdu.ac.lk/handle/345/5300 2025-09-02T09:28:15Z 2022-01-10T00:00:00Z

Sher-locked: a Hybrid Deep Learning Model Based Mobile Platform for Social Media Fact-checking Goonathilake, MDPP; Kumara, PPNV In the present context, false news can be easily constructed and circulated through various social media platforms. As a result, people on those platforms have difficulty in distinguishing between correct and incorrect information. Therefore, a firm desire appears to develop a fact-checking platform to address this issue. From this research study, the authors present ‘Sher-Locked’ which is a hybrid deep learning model based mobile platform to fact-check information on social media. The process of checking and verifying information is referred to as fact-checking. A hybrid deep learning model which is mainly focused on CNN and RNN-LSTM networks integrated with the mobile application to check and verify information on social media. The high-level characteristics and interdependencies among the input text capture from the hybrid model. The mobile application consists of several features such as fact-checking, daily news updates, news reporting, social media trends and daily COVID-19 reports. Flutter chose as the mobile application development framework along with Firebase as the backend development framework with REST APIs to develop the entire system. When checking and verifying the information mitigating on social media, the hybrid model achieved a 92% accuracy by surpassing most of the traditional models today with 91% score rates for Precision, Recall and F1-Score. After delivering the mobile app as a complete system to various users for testing, the authors discovered that the user satisfaction and usability rates are high when compared to other related software.

2022-01-10T00:00:00Z Baddegamage, SI De Silva, LNC Goonethilake, MDJS https://ir.kdu.ac.lk/handle/345/5299 2025-09-02T09:29:13Z 2022-01-10T00:00:00Z

Information Management for Sri Lankan Vegetable Farmers: Effectiveness of ICT Applications Baddegamage, SI; De Silva, LNC; Goonethilake, MDJS The scenario behind Sri Lankan agriculture sector is that one-third of the country’s population who engage in agriculture contributes only 7% of the GDP. The distribution of smaller amounts of income among large communities increases poverty among farmers in Sri Lanka. This limited income shrinks further due to sudden price drops, wastage, damages and oversupply. Various types of ICT-based solutions have been provided to eliminate poverty among farmers in Sri Lanka. However, research findings and literature show that most farmers are still suffering in poverty in an age of information even with the availability of many forms of information sources required for farmers. Due to some issues or reasons, farmers do not continuously use information systems and available information systems become obsolete within a short period due to lack of continual use. The research explores reasons for the limited use of information and communication technology-based agricultural information systems among Sri Lankan farming community. The research collected data using literature review, questionnaires and interviews from 76 farmers in four districts of Sri Lanka. Weekly average prices of three selected vegetables and selling offers received for a digital classified AgriApp were observed for one year and collected data was analyzed to identify farmers’ and market behavior patterns. Research findings will help to increase ICT practices in agriculture, reduce wastage, control price fluctuation, and eliminate oversupply. It will ensure a continuous supply of vegetables and food security of the nation.

2022-01-10T00:00:00Z Nanayakkara, VSS Wanniarachchi, WAAM Vidanagama, DU https://ir.kdu.ac.lk/handle/345/5298 2025-09-16T09:22:50Z 2022-01-10T00:00:00Z

Investor Driven Adaptive and Automated Stock Market Portfolio Management Platform with Stock Prices Prediction for Colombo Stock Exchange of Sri Lanka Nanayakkara, VSS; Wanniarachchi, WAAM; Vidanagama, DU Over the past few years various studies have been conducted to develop an optimum stock market related portfolio management platform that will assist investors to actively perform the portfolio management process. Risk and level of investor participation is considered to be one of the challenging aspects identified for optimum portfolio management. Along with portfolio management, stock price prediction is one of the key contributing factors that helps an investor to make mid and long-term strategic investment decisions. Various concepts are evaluated and studied thoroughly to determine the most accurate algorithm to implement a stock price-based prediction system. Currently, Colombo Stock Exchange have identified a desperate requirement of a portfolio management system with prediction capabilities to support the local and foreign investors to actively engage in trading activities in different stock exchanges in different countries. A critical study has been conducted using supportive research papers, studying similar applications which are developed so far and using various requirement elicitation techniques to determine the functional requirements, non-functional requirements, investor requirements and User Interface/User Experience (UI/UX) considerations. The paper further describes various technological mechanisms implemented and system architectures used to develop the portfolio management and stock price prediction system. Accordingly, the implementation of Brownian Motion algorithm-based model and LSTM (Long Short-Term Memory) model are presented in detail by the author. Finally, evaluation and testing results of the completed system and stock price prediction models are presented to prove the successfulness of the completed application and accuracy of the models implemented

2022-01-10T00:00:00Z