Tuning Hyperparameters of Classification Algorithms in Artificial Intelligence and Machine Learning

Abstract

This article applies modern approaches of artificial intelligence and machine learning technologies for the early detection and diagnosis of chronic kidney disease based on clinical data. The study analyzes the application of Support Vector Machine, Logistic Regression, and K-nearest neighbor algorithms to clinical data, the optimization of hyperparameters, and the evaluation metrics of these approaches. The work was carried out in the following order: first, the dataset was analyzed and missing values were filled in. Then, standardization and normalization were performed, followed by the application of categorical column encoding and train/test split methods. For each of the selected algorithms, optimal hyperparameters were first chosen using default settings, and then through Grid Search and Random Search methods. The results were subsequently evaluated based on cross-validation. The evaluation metrics were expressed through accuracy, precision, recall, F1-score, ROC curve, and confusion matrix. The research results demonstrated that the effectiveness of the outcomes was achieved through optimizing hyperparameters in the SVM algorithm. The model achieved an increase in generalizability and clinical reliability. In the KNN model, parameter tuning also improved the model's performance, while Logistic regression demonstrated simplicity and stability. The research results demonstrated the importance of utilizing artificial intelligence in medical diagnostics and clinical decision-making, as well as highlighting the significance of modern algorithms.

Ähnliche Arbeiten