Optimizing Extreme Learning Machine (ELM) Input Weights and Biases Using the Bat Algorithm (BA)

In this paper, we implemented the Bat Algorithm (BA) to optimize the Extreme Learning Machine (ELM) model. The optimization primarily targeted the input weights and bias parameters, where BA's frequency tuning and loudness adjustment mechanisms were coded to dynamically explore optimal parameter combinations. This approach resulted in a marked improvement in diagnostic accuracy. Through data analysis, we discovered previously unreported features with high diagnostic value, identified via feature importance scoring algorithms. Additionally, we compared our method against conventional approaches using standard evaluation metrics (e.g., accuracy, F1-score), demonstrating superior performance in diagnostic precision. The implementation included cross-validation protocols and statistical significance testing to ensure robustness. In conclusion, our research introduces novel methodologies for medical diagnosis by integrating swarm intelligence with neural network optimization.