Performance Optimization of Machine Learning Algorithms by using Crayfish Optimizer

The need for optimization in machine learning algorithms is essential to enhance prediction accuracy and efficiency across various applications. The Crayfish Optimization Algorithm addresses parameter tuning and feature selection challenges in Linear Regression, Support Vector Machine, K-nearest neighbors, and Isolation Forest in this work. The study uses four datasets representing different domains ranging from medical domain to banking sector. The Crayfish Optimization Algorithm is incorporated into the training process to optimize hyperparameters and reduce feature redundancy. For the Wisconsin Breast Cancer Dataset, the optimized SVM model achieves an accuracy of 97.1%, while the optimized LR model attains 98.07% accuracy. In the case of the Credit Card Fraud Detection Dataset, the optimized KNN model reaches 99.92% accuracy, and the optimized DT model achieves 97% accuracy. For the Glaucoma Image Dataset, the optimized SVM model exhibits a significant increase in precision by 5%, attaining the highest precision value of 95%. The optimized SVM and LR models demonstrate lower mean squared error scores in the Stock Market Dataset. These findings highlight the potential of the COA in enhancing the reliability and adaptability of machine learning models across diverse domains.