Triple-Mutation Bat Algorithm–Optimized Extreme Learning Machine for Fetal Health Classification

Abstract

Fetal health assessment is crucial for preventing perinatal risks; however, the manual interpretation of cardiotocography (CTG) signals remains susceptible to variability and diagnostic delays. To address this challenge, this study aims to develop an accurate and computationally efficient model for automated fetal health classification. This study proposes a hybrid intelligent model named TMBA–ELM (Extreme Learning Machine parameters optimized with Triple Mutation Bat Algorithm) for accurate and efficient classification of fetal health conditions. The purpose of this research is to improve the diagnostic reliability of CTG data analysis while maintaining low computational complexity. The proposed TMBA–ELM integrates the adaptive exploration–exploitation mechanism of the Tuned Modified Bat Algorithm (BA) with the fast learning capability of the Extreme Learning Machine (ELM) to optimize both the hidden neuron size and the activation function. Three adaptive mutation strategies: Cauchy, Gaussian, and position modification—are applied to enhance convergence and population diversity. The model was evaluated using an imbalanced CTG dataset containing 2,126 samples classified as Normal, Suspect, or Pathological, and benchmarked against BA-ELM (ELM parameters tunned with BA), EMD-FA-ELM (ELM parameters tunned with firefly algorithm and the data decompositioned by empirical mode decomposition), and PSO-EM-ELM (error minimized ELM parameters tunned with particle swarm optimization) using five performance metrics: accuracy, precision, recall, F1-score, and computation time. Experimental results show that TMBA–ELM achieve high performance with 91.4% accuracy, 82.8% precision, 77.77% recall, and 79.93% F1-score, while maintaining the high computation time (164.23 second) with moderate stability (deviation = 12.76 second). The model also demonstrates improved recall for minority classes, confirming robustness in handling data imbalance. Although have bad computational efficiency, TMBA-ELM offers highest recall and F1-score, making it suitable for real-time fetal monitoring. The originality of this study lies in integrating triple adaptive mutation strategies within the Bat Algorithm to optimize ELM parameters.