blend_ensemble.py 文件源码

def online(X_org, y_org, test_x, test_uid):
    n_folds = 5
    verbose = True
    shuffle = False

    X = X_org
    y = y_org
    X_submission = test_x

    if shuffle:
        idx = np.random.permutation(y.size)
        X = X[idx]
        y = y[idx]

    skf = list(StratifiedKFold(y, n_folds))

    clfs = [
        RandomForestClassifier().set_params(**INITIAL_PARAMS.get("RFC:one", {})),
        ExtraTreesClassifier().set_params(**INITIAL_PARAMS.get("ETC:one", {})),
        GradientBoostingClassifier().set_params(**INITIAL_PARAMS.get("GBC:one", {})),
        LogisticRegression().set_params(**INITIAL_PARAMS.get("LR:one", {})),
        xgb.XGBClassifier().set_params(**INITIAL_PARAMS.get("XGBC:two", {})),
        xgb.XGBClassifier().set_params(**INITIAL_PARAMS.get("XGBC:one", {})),
        ]

    print "Creating train and test sets for blending."

    dataset_blend_train = np.zeros((X.shape[0], len(clfs)))
    dataset_blend_test = np.zeros((X_submission.shape[0], len(clfs)))

    for j, clf in enumerate(clfs):
        print j, clf
        dataset_blend_test_j = np.zeros((X_submission.shape[0], len(skf)))
        for i, (train, test) in enumerate(skf):
            print "Fold", i
            X_train = X[train]
            y_train = y[train]
            X_test = X[test]
            y_test = y[test]
            clf.fit(X_train, y_train)
            y_submission = clf.predict_proba(X_test)[:,1]
            dataset_blend_train[test, j] = y_submission
            dataset_blend_test_j[:, i] = clf.predict_proba(X_submission)[:,1]
        dataset_blend_test[:,j] = dataset_blend_test_j.mean(1)

    print "Blending."
    # clf = LogisticRegression(C=2, penalty='l2', class_weight='balanced', n_jobs=-1)
    clf = linear_model.RidgeCV(
            alphas=np.linspace(0, 200), cv=LM_CV_NUM)
    # clf = GradientBoostingClassifier(learning_rate=0.02, subsample=0.5, max_depth=6, n_estimators=100)
    clf.fit(dataset_blend_train, y)
    # y_submission = clf.predict_proba(dataset_blend_test)[:,1]
    print clf.coef_, clf.intercept_
    y_submission = clf.predict(dataset_blend_test)  # for RidgeCV

    print "Linear stretch of predictions to [0,1]"
    y_submission = (y_submission - y_submission.min()) / (y_submission.max() - y_submission.min())
    print "blend result"
    save_submission(os.path.join(consts.SUBMISSION_PATH,
                                     MODEL_NAME + '_' + strftime("%m_%d_%H_%M_%S", localtime()) + '.csv'),
                        test_uid, y_submission)