K-mer Features
No single stable selected k-mer subset appeared across all folds. GA feature selection reduced dimensionality but did not discover a robust disease-discriminative signal. The k-mer representation used relative frequencies of all 400 possib…
1 sources - 5 claims
No single stable selected k-mer subset appeared across all folds. GA feature selection reduced dimensionality but did not discover a robust disease-discriminative signal. The k-mer representation used relative frequencies of all 400 possible dipeptides. The genetic algorithm selected fold-specific k-mer subsets averaging about 84 features. K-mer and hybrid representations often achieved higher F1 by over-predicting the positive class.