feature_selection.py

def select(X):
    selects = []
    
    selector = SelectKBest(chi2, k='all').fit(X, y)
    scores = selector.scores_
        
    q3 = np.quantile(scores, 0.75)
    q1 = np.quantile(scores, 0.25)
    iqr = q3 - q1
    threshold = q3 + 1.5 * iqr

    for col, val in zip(X.columns, scores):
        cond1 = val > threshold
        cond2 = col in test_hot.columns or col in test.columns
        
        if cond1 and cond2:
            if col not in test_hot.columns:
                test_hot.loc[:, col] = 0
            selects.append(col)
        
    return selects