Journal of Knowledge in Data Science and Information Management

Explainable Artificial Intelligence for Tuberculosis Detection: A Comprehensive Review of Techniques, Challenges, and Future Directions

Kangana Soni — 2026-05-02

Explainable artificial intelligence (XAI) has appeared as an important study domain in healthcare, resolving the constraints of black-box machine learning and deep learning frameworks. In the field of tuberculosis recognition, AI methods have proven precision in evaluating chest X-ray scans and medical data; however, the absence of explainability restricts their implementation in healthcare settings. This study shows an extensive analysis of 50 research papers emphasizing on the utilization of XAI in tuberculosis recognition. The chosen papers are evaluated based on techniques, datasets, interpretability methods and assessment parameters. Frequently applied XAI procedures, such as LIME, SHAP, and Grad-CAM, are studied in depth. The analysis shows that although these methods increase transparency, they are frequently restricted to certain kinds of descriptions, such as local-global or spatial, causing partial awareness of system performance. Additionally, this paper detects important problems, including the lack of consistent analysis parameters, restricted application of hybrid XAI systems and absence of medical confirmation. The results indicate that combining multiple XAI methods and implementing measurable assessment procedures can considerably improve the trustworthiness and reliability of AI-driven TB identification frameworks. Ultimately, upcoming study areas are described to lead the advancement of clear, understandable and medically usable AI algorithms in healthcare.

AI-based Intelligent Smart City Complaint Management System

Narige Jyoshna — 2026-06-01

As urbanization increases, the complexities faced while handling citizen problems and complaints increase as well. Complaint management systems in smart cities rely on artificial intelligence (AI). This project revolves around a complaint management system for smart cities that aims to streamline the process by reducing human intervention while handling complaints effectively and accurately. The system is fully automated; complaints are raised by citizens, and then AI processes them automatically by prioritizing, deduplicating, classifying and tracking them. The system was coded in Python language using Tkinter library for its user interface design, whereas for record-keeping and storage of information, SQLite database is used. Complaints management involves the implementation of natural language processing techniques to analyze complaints received in the form of texts. For classification of complaints, the text-to-vector conversion technique called term frequency-inverse document frequency (TF-IDF) and Naive Bayesian classifier were applied. Furthermore, cosine similarity technique was implemented to detect duplicate complaints. Moreover, a priority system was developed that ranks complaints in the order of urgency, so complaints with high priority come first and get processed. When a complaint is raised, a tracking ID is provided to the complainant so he/she can track its progress. All this will ensure that the reaction time is reduced while improving accuracy, thus making the entire process easier and better scalable. Ultimately, it can be argued that this is an intelligent solution to problems of smart cities.

Edge versus Cloud: Evaluating Big Data Processing Paradigms for IoT Applications

Shivang Mishra — 2026-05-13

The explosive growth of Internet of Things (IoT) ecosystems across industries, from factory floors and hospital wards to smart farms and urban infrastructure, has fundamentally changed how data processing is perceived. Billions of connected devices now generate continuous streams of sensor data, telemetry, images, and events, creating a need for computing architectures that are fast, scalable, secure, and cost-effective. Two paradigms dominate today's deployments: edge computing, which processes data close to where it is generated, and cloud computing, which centralises massive computational power in global data centres. This study offers a structured comparison of both paradigms in the context of IoT big data. The study introduce a two-category taxonomy: Type 1 workloads requiring millisecond-level responses, and Type 2 workloads suited for large-scale batch analytics and argue that the optimal architecture depends heavily on the workload’s latency profile, privacy requirements, scale, and operational context. The analysis covers architecture, performance, security, cost, and governance dimensions, and is grounded in real-world case studies across industrial automation, smart cities, precision agriculture, and healthcare. The study concludes that a thoughtfully designed hybrid architecture combining edge autonomy with cloud depth is the most effective path forward for most production IoT systems.

Digital Preservation of Ethno-medicinal Knowledge: Oral Traditions of the Tharu Tribe in Lakhimpur Kheri, Uttar Pradesh—A Library and Information Science (LIS) Framework

Karan Jaiswal — 2026-05-14

The Ethno-medicinal knowledge of the Tharu tribe in Lakhimpur Kheri is threatened by rapid urbanization. This study proposes an LIS-based digital institutional repository (DIR) to document their oral traditions through field surveys and interviews. By leveraging this framework, the research aims to preserve indigenous heritage for future pharmacological study and community identity. Key components of the framework include the use of multimedia documentation (audio/video), the implementation of knowledge management (KM) techniques, and addressing ethical concerns regarding intellectual property rights (IPR). The findings suggest that a structured LIS intervention can transform oral “hidden” knowledge into a “visible” digital resource, fostering community identity and providing a valuable database for pharmacological research.