-
Performance Analysis of Experimental Data for Heart Disease Prediction Using Unsupervised K-Means Clustering
- Section
- Articles
- Published
- Pages
- 1
Abstract
Background: Heart disease is a leading cause of death worldwide, and medical databases hold large volumes of data from which diagnostic patterns could be extracted. This study develops a predictive system for heart disease using unsupervised K-means clustering and examines how the number of clusters (k) affects performance.
Methods: A heart disease dataset from a hospital in Srinagar, Jammu and Kashmir, was pre-processed using information-gain attribute evaluation with a ranker search. Redundant, missing and erroneous instances were removed, leaving 200 instances with 12 attributes. K-means models were built in Weka with k = 2, 4, 8 and 14. They were compared on squared error, time taken and number of iterations.
Results: Squared error fell as k increased, from 27.45% at k = 2 to 23.09% at k = 4, 14.23% at k = 8 and 3.04% at k = 14. Over the same range, build time rose from 0.0 to 0.05 seconds and the number of iterations from 3 to 9.
Conclusion: Larger numbers of clusters reduced error but increased computation time, iterations and model complexity. The eight-cluster model (k = 8) was selected as the most appropriate. A moderate number of clusters therefore appears best suited to heart disease prediction, and the optimal k depends on dataset size.
Keywords Performance Analysis Experimental Data Heart Disease Prediction Unsupervised K-Means Clustering
Fulltext PDF -
Smart Civic Connect: A Role-Based Web Platform for Reporting, Assigning and Tracking Urban Infrastructure Complaints
- Section
- Articles
- Published
- Pages
- 1–6
Abstract
Potholes, overflowing garbage, dead streetlights and burst water mains are reported to local bodies every day, yet citizens rarely learn what happened to their complaint. The problem is less a shortage of repair crews than the lack of a shared, trackable channel between the three parties involved: the citizen who sees the defect, the administrator who decides who should fix it, and the worker who fixes it. This paper presents Smart Civic Connect, a full-stack web platform that connects citizens, municipal administrators and maintenance workers in one workflow. It is built with Python and Django, with Bootstrap, Leaflet.js and Chart.js on the client and SQLite or PostgreSQL for storage. It introduces a structured complaint model with ten civic categories, four priority levels, map coordinates and photographs; a seven-state lifecycle with a complete history of every transition; server-enforced role-based access; evidence-based resolution through a completion photograph; a public community feed and colour-coded map; an in-app notification service with a fail-safe e-mail fallback; and an administrative analytics dashboard with CSV export. We describe the architecture and implementation, summarise the automated test suite, and discuss limitations and future work.
Keywords Civic Complaint Management E-Governance Smart City Django Role-Based Access Control Geospatial Mapping Citizen Participation
PDF