Скачать или смотреть Statistical Machine Learning on DNA Microarrays: Model Comparison and Interpretation

This video presents a practical statistical machine learning analysis of a DNA microarray prostate cancer dataset. The data consist of a small number of samples and a large number of gene expression variables, making the problem inherently high-dimensional.

I discuss the structure of the dataset, univariate screening using the Kruskal–Wallis test, and the behavior of classification trees and k-nearest neighbor classifiers. Model performance is compared using ROC curves and repeated train–test splits, followed by an analysis of variance (ANOVA) to assess differences in mean test error across learning machines.

The focus of the presentation is on understanding why the models behave as they do, rather than only reporting performance metrics. The results highlight the importance of variable relevance, model complexity, and careful evaluation in high-dimensional statistical learning problems.