Early prediction of gallstone disease with various feature selection methods from bioimpedance and laboratory data

Değirmenci Akdeniz M.

International Conference on Next-Generation Engineering Technologies (ISNET'2025), Antalya, Turkey, 30 October - 01 November 2025, pp.60-61, (Full Text)

  • Publication Type: Conference Paper / Full Text
  • City: Antalya
  • Country: Turkey
  • Page Numbers: pp.60-61
  • Kütahya Health Sciences University Affiliated: Yes

Abstract

Machine learning-based approaches can be used for early prediction of gallstone disease. This study investigates the effect of five different feature selection methods to reduce features and classify bioimpedance and laboratory data by comparing lower-dimensional matrices using well-known classifier algorithms. A publicly available dataset, which consist of 319 samples with 38 features, is used in analysis. The reduced feature sets were obtained using Principal Component Analysis (PCA), One-Way Variance Anaylsis (ANOVA), Minimum Redundancy Maximum Relevance (mRMR), Kruskal-Wallis (KW), and Chi-Square test (X2) methods. Nine distinct classifiers are used to classify all feature sets. A maximum accuracy of 90.48% was obtained using X2 method and Efficient Linear Support Vector Machine classifier.