Порівняльний аналіз алгоритмів кластеризації та методи оптимізації логістичної регресії в задачах медичної діагностики

Abstract

The paper presents a comparative analysis of clustering methods and optimization approaches for classification models based on the Breast Cancer Wisconsin dataset. K-Means, Spectral Clustering, and Gaussian Mixture methods were applied for clustering, along with Principal Component Analysis (PCA) for dimensionality reduction. Classification models based on logistic regression with parameter optimization via gradient descent and genetic algorithm were built and compared. It was established that the highest classification efficiency is achieved by sklearn's logistic regression.