Recently to get best features I have used SelectKBest(score_func=, k=20). f_classif computes anova f-value between feature for classification tasks. I have used it and got best results. I learnt anova f-test computes ratio of 'between class variance' to 'within class variance'. But I have following questions:
1) Does f_classif use combination of features to give f-score ?. 2) Can I have a pseudo code of how fit function of SelectKBest works ?. 3) How does f_classif work in sklearn ?. Thanks in advance.