Conference papers

Permanent URI for this communityhttps://rda.sliit.lk/handle/123456789/4889

Browse

Filters

2025 - 20264

Advanced Search

Filter by

Settings

Subject: search.filters.subject.explainable AI

Search Results

Now showing 1 - 4 of 4
  • Thumbnail Image
    ItemEmbargo
    Explainable Multilingual Sentiment Analysis for Sinhala, English and Code-Mixed Banking Reviews
    (Institute of Electrical and Electronics Engineers Inc., 2025-12-09) Senevirathna, P; Adhikari, N; Navojith, T; Rizvi, A; Kasthurirathna, D; Abeywardhana, L
    The growth of multilingual and code-mixed communication in the financial sector presents significant challenges for sentiment analysis, particularly in low-resource languages such as Sinhala. Existing NLP models often underperform due to limited annotated data, domain-specific vocabulary gaps, and a lack of transparency. This paper presents a multilingual and explainable sentiment analysis framework that addresses the challenges of low-resource and code-mixed languages by analyzing banking customer feedback in English, Sinhala, and Sinhala-English code-mixed formats. A dataset of 13,000 aspect-tagged banking reviews spanning five key banking aspects is used for evaluation. The framework classifies comments into positive, neutral, or negative sentiment and provides interpretable explanations using SHAP (Shapley Additive Explanations) and LIME (Local Interpretable Model Agnostic Explanations). English sentiment classification is performed using a fine-tuned BERT-base-uncased model, while Sinhala/code-mixed sentiment is handled by a hybrid XLM RoBERTa with a domain-specific lexicon correction approach. The framework achieves 92.3% accuracy and 0.89 F1-score for English, and 88.4% accuracy and 0.84 F1-score for Sinhala/code-mixed data. The proposed system addresses reproducibility, interpretability, and domain adaptation gaps, offering a deployable solution for multilingual financial sentiment monitoring.
  • Thumbnail Image
    ItemEmbargo
    Multimodal AI Framework for Personalized and Context-Aware Skin Disease Diagnosis, Monitoring, and Treatment Support
    (Institute of Electrical and Electronics Engineers, 2026-05-22) Wijesinghe H.W.M.O.P.; Laksopan R; Mihisandali W.K.M.; Devindi K.P.T.; Weerasinghe, L; De Silva, A
    Dermatoscopic assessment of skin diseases based on visual morphology may not provide sufficient discrimination due to differences in cutaneous appearance, the severity of disease symptoms and individual biological or environmental factors. Current artificial intelligence (AI)-based dermatological systems mainly integrate unimodal image-based data which is constrained by comparative diagnostic performance in visually ambiguous conditions and across different skin complexions. Image-only approaches also do not utilize patient-reported symptoms needed to tailor treatment plans. In this study, we explore the potential of a unified multimodal AI framework towards robust, context-aware and patient- centric skin disease diagnosis, monitoring and treatment support. The framework combines deep learning-based image analysis with symptom-aware inputs extracted from voice recordings and structured text, which allows for improved diagnostic reliability. The proposed framework also introduces an explainable severity assessment module which evaluates disease progression via interpretable features and rule-based score. Domain adaptation methods further employed lead to better generalization for out-of-distribution data originating from different populations and reduce model bias. A knowledge-driven recommendation module generates context-aware personalized treatment recommendations based on predicted disease categories and patient-related information. Experimental results demonstrate that the proposed multimodal framework improves contextual understanding and robustness in visually ambiguous cases while enhancing interpretability, improved generalization, and practical applicability in teledermatology environments, while adding interpretability, fairness and real-world applicability of teledermatology systems.
  • Thumbnail Image
    ItemEmbargo
    A Reinforcement Learning Approach with Human in the Loop to Explainable Insurance Risk Scoring and Intelligent Policy Portfolio Optimization
    (Institute of Electrical and Electronics Engineers, 2026-05-29) Gamage, C; Kasthuriarachchi, T; Denuwan, C; Mallawaarachchi, P; Abeywardhana, L; Nawarathne, M
    Assessing individual risk accurately and optimizing insurance portfolios in real time remain major challenges due to static actuarial tables, opaque models, and fragmented analytical pipelines. This paper proposes a two-part Explainable AI (XAI) framework addressing both issues. The first component, Artificial Intelligence-driven risk scoring with human-in-the-loop (HIL) weight adjustment, uses a Proximal Policy Optimization (PPO) agent to suggest feature-based changes to an insurer's risk-equation weights. Shapley Additive Explanations(SHAP) attributions and Generative AI reasoning make these changes interpretable, allowing human reviewers to approve modifications that are immediately applied to future customers, creating a self-improving loop. The second component, AI-driven policy optimization, leverages a PPO supported by an XGBoost expense regressor, SHAP/LIME explainability, PPO agent and a Retrieval-Augmented Generation (RAG) layer for rider assignment. Both components share a data backbone of 100,000 anonymized insurance records stored in MongoDB, enabling incremental updates without reprocessing. Experiments show the XGBoost regressor achieves Root Mean Square Error (RMSE) 0.4406 and Mean Absolute Error (MAE) 0.3600, the HIL guided agent increases average episodic reward by 10.3%, and the RAG layer reaches 91.7% rider-assignment accuracy. The framework significantly enhances predictive accuracy, interpretability, regulatory traceability, and portfolio adaptability compared to traditional actuarial and black-box approaches.
  • Thumbnail Image
    ItemEmbargo
    An Explainable Deep Learning Framework for Coconut Disease Detection Using MobileNetV2, Super-Resolution, and Grad-CAM++
    (Institute of Electrical and Electronics Engineers Inc., 2025) Balasooriya R.C.; Adithya E.L.A.Y; Gunarathne M.M.S.U; Silva T.C.D; Lokuliyana, S; Wijesiri, P
    Coconut production is a significant industry in Sri Lanka's economy and food security. However, it is constantly under threat from diseases such as Grey Leaf Spot and pests such as Coconut Mites (Aceria guerreronis). Detection must be early, but it is difficult, especially in field conditions where image quality is low and symptoms are not visually distinguishable. This paper proposes a two-stage deep learning solution to enhance and automate disease and pest recognition with a lightweight and mobile system. The system combines Real-ESRGAN based image super-resolution to restore visual detail in poor-quality mobile images and MobileNetV2-based classification, a lightweight convolutional neural network. The model recognizes grey leaf spot with over 97% accuracy and greatly enhanced mite recognition performance when combined with super-resolution preprocessing. In the interest of transparency and trust for users, the Grad-CAM++ and LIME interpretation techniques are utilized, and visual explanations of the predictions are presented. A mobile application was created with React Native and integrated with a Flask-based backend to enable real-time image enhancement and classification to facilitate practical deployment. Smartphone-captured field-level photos were preprocessed and categorized into healthy, diseased, and non-coconut samples. Farmers can use the proposed system in real time because it maintains good accuracy while being computationally efficient. This framework provides a scalable method for intelligent and sustainable agriculture.