Title: Retrieval optimisation in case-based reasoning systems applied to healthcare

Authors: Seema Sharma; Deepti Mehrotra; Narjès Bellamine Ben Saoud

Addresses: Amity University, Uttar Pradesh, Amity Rd., Sector 125, Noida, Uttar Pradesh 201313, India ' Jaypee Institute Information Technology, Noida, India ' ENSI, University of Manouba Tunisia, Campus Universitaire de La Manouba, Cité Nasr 2010, Tunisia

Abstract: Case-based reasoning (CBR) is a cognitive approach that solves new problems by referencing solutions from similar past cases. It is particularly effective in domains where knowledge is incomplete and exceptions are common, such as medical decision-making. This study focuses on enhancing the retrieval phase of CBR systems that manage large-scale medical case bases. To achieve this, various k-nearest neighbour (KNN) algorithm variants were evaluated to identify the most suitable for medical applications. The study compared standard KNN, fuzzy KNN (F-KNN), fuzzy k-nearest centroid neighbour (F-KNCN), Bonferroni mean-based fuzzy KNN (BMF-KNN), Bonferroni mean-based fuzzy KNCN (BMF-KNCN), and random KNN (R-KNN) algorithms. These were tested on four benchmark heart disease datasets within a CBR framework. The results demonstrate that R-KNN achieves the highest accuracy of 92%, outperforming the other variants. Thus, R-KNN is identified as the most effective retrieval method for improving medical CBR systems.

Keywords: case-based reasoning; CBR; random k-nearest neighbours algorithm; k-nearest neighbour; K-NN; fuzzy k-nearest neighbour; F-KNN; fuzzy k-nearest centroid neighbour; F-KNCN; Bonferroni mean-based fuzzy k-nearest neighbour; BMF-KNN; Bonferroni mean-based fuzzy k-nearest centroid neighbour; BMF-KNCN.

DOI: 10.1504/IJEH.2026.156150

International Journal of Electronic Healthcare, 2026 Vol.14 No.3, pp.189 - 215

Received: 05 Jun 2023
Accepted: 10 Apr 2024
Published online: 07 Sep 2026 *

Full-text access for editors Full-text access for subscribers Purchase this article Comment on this article