http://repo.lib.sab.ac.lk:8080/xmlui/handle/123456789/19 Sat, 06 Jun 2026 12:00:44 GMT 2026-06-06T12:00:44Z http://repo.lib.sab.ac.lk:8080/xmlui/handle/susl/5333 A Hybrid LLM-Driven Recommendation System for Selecting Flutter Packages Using Structured Metadata and Stack Overflow Knowledge Bandara, R.M.I.M.; Chathumini, K.G.L.; Kumara, B.T.G.S. The rapid growth of the Flutter ecosystem has built significant challenges for the developers in choosing the right third-party packages. Existing popularity-based methods fail to capture developer intent or semantic alignment with real-world requirements. To address this gap, this research presents a hybrid LLM-based recommendation system that integrates structured metadata from Flutter repositories with unstructured knowledge from Stack Overflow. With three LLMs comparatively evaluated, GPT-4- Turbo was chosen based on the accuracy, consistency, and low hallucination rate. The final recommendation model uses a weighted scoring method that integrates popularity, recency, likes, downloads, accepted answers and LLM-based semantic relevance, which generates ranked package recommendations based on the user’s requirements. Experimental evaluation across multiple test scenarios achieved a precision of 93.2%, recall of 91.5%, F1-score of 92.3%, and an average semantic relevance score of 94/100. Expert evaluation by Flutter practitioners with six to ten years of experience confirmed that 92% of the recommendations aligned with realworld development requirements. Even though the paper focuses on state management packages, the approach methodology can be generalized to other Flutter package categories. The results highlight the potential of LLMbased hybrid systems to improve developer decision-making. Wed, 28 Jan 2026 00:00:00 GMT http://repo.lib.sab.ac.lk:8080/xmlui/handle/susl/5333 2026-01-28T00:00:00Z http://repo.lib.sab.ac.lk:8080/xmlui/handle/susl/5332 AI-Driven Risk Prediction in Software Development Environment Sanchayan, R.; Somaweera, W.T.S.; Sandaruwan, R.M.T. Software development projects are constantly carried out in circumstances that are highly unexpected and dynamic. Unpredictability caused by frequent demand changes, shifting responsibilities, and delivery schedule concerns can all have a negative impact on the process. The implementation of traditional risk management approaches can be quite challenging, and in many situations, manual processes rely on periodic assessments. This research aims to establish machine learning as a unique approach for prediction and decision-making in software project management, with the goal of overcoming the limits of existing methods. The research was built on a thorough collection and investigation of past project data. Before evaluating several machine learning algorithms on structured project datasets, the data was thoroughly preprocessed and feature extracted. XGBoost was the most effective model, capturing the complicated nonlinear relationships between many project variables and identifying risk indicators. The model achieved 83.54% accuracy, 84.19% precision, 83.54% recall, and a low F1-score of 79.49%, indicating that it is an acceptable option for predictive risk analysis in real-world applications. The data from the current project, which was used in short-term live validation findings, show that the model can detect high-risk processes faster than manual evaluation approaches. This allows project managers to take rapid action, such as altering resource allocation, revising task orders, and addressing developing difficulties. The study states that using machine learning techniques improves the accuracy, speed, and reliability of software project risk assessment while also transforming current project management practices through large-scale data-driven advancements and the integration of intelligent, automated monitoring systems. Wed, 28 Jan 2026 00:00:00 GMT http://repo.lib.sab.ac.lk:8080/xmlui/handle/susl/5332 2026-01-28T00:00:00Z http://repo.lib.sab.ac.lk:8080/xmlui/handle/susl/5331 AI Based Flutter Interface Generator Using Natural Language Commands with Social and Ethical Considerations Andarawewa, K.M.; Chathumini, K.G.L. The accelerated pace of development of Artificial Intelligence (AI) and the emerging opportunities to automate the generation of user interfaces (UI) are explored in this research, especially within the context of cross-platform development tools like the Flutter platform. This research aimed to develop customed and open access LLM models, to create User Interfaces (UI) effectively. Then, the social and ethical implications of AI based UI design creation were assessed through a questionnaire survey. The new framework combines all free-tier Large Language Models (Gemini, Groq, Cohere, Hugging Face, Open Router) with individually developed models created viaMPNet-Base sentence transformers. The individually developed model performed training on newly developed data sets of 10,000 samples, defining varied UI designs such as authentication pages, dashboards, form elements, and e-commerce UI designs. The new framework allows real-time generation of UI source code for both texts and voice commands developed via Flutter and Python FastAPI development tools. The methodology to evaluate technological advancements utilized performance assessment and conducting surveys to measure UI source-code accuracy and related social-ethical perceptions from 220 IT professionals. The new framework resulted in 87.01% accuracy and 93.31% F1 scores. Analysis disclosed major defects such as pattern repetition, color contrast issues, and incongruent points as significantly prioritized among IT professionals. The new framework showed technological feasibility along with justified needs to introduce social and ethical considerations to ensure greater monitoring and human control in generating UI via AIs. Wed, 28 Jan 2026 00:00:00 GMT http://repo.lib.sab.ac.lk:8080/xmlui/handle/susl/5331 2026-01-28T00:00:00Z http://repo.lib.sab.ac.lk:8080/xmlui/handle/susl/5330 Detecting Developer Burnout Through Sentiment Trends in Software Repositories Using DistilBERT Samaraweera, S.P.I.L.D.; Adeeba, S. Recent studies have indicated that severe chronic stress relates to burnout, and Software Engineering (SE) is no exception, having greater than 57chronic stress. Conventional methods of evaluation employ surveys and biometric measurements, which are extremely intrusive. In the domain of remote working teams, continual real-time measurement lacks feasibility. This study investigates the feasibility of reliably detecting early signs of developer burnout using a sentiment-focused framework, unobtrusively. Using the 20 Years of Issues and Commits of Mozilla and Apache Development (20-MAD), the dataset of 2,166,239 messages of communication (2,390 developers) is processed using a negation-aware preprocessing technique, DistilBERT sentiment-analysis framework, and Zscore temporal analysis to quantify and capture burnout. Findings highlight the detection of weekly and monthly burnout signals with a weighted classification of 84.09% precision, traditional models vs. DistilBERT. Statistically significant differences in sentiment, with a significant effect of true vs classification, were demonstrated. 462 developers demonstrated evidence of mild stress, 3 severe, and 24 moderate burnouts, during the week. The latter, independently, demonstrated evidence of sustained burnout across the period, which was confirmed by the monthly analysis of 13 severe moderate chronic stress cases and 24 sustained burnout cases. Integration based on GitHub has been made available for the real-time monitoring of burnout using advanced machine learning methods, which allow for high-speed inference within 100ms on a message to provide privacy to the user. The framework permits monitoring the well-being of developers, which consequently gives organizations the ability to run preventive programs. Wed, 28 Jan 2026 00:00:00 GMT http://repo.lib.sab.ac.lk:8080/xmlui/handle/susl/5330 2026-01-28T00:00:00Z