single_feature.py 文件源码

def train_test(X, Y, ratio):
    estimators = build_model_mlp()
    clf = Pipeline(estimators)
    # clf = RandomForestClassifier(n_jobs=-1, n_estimators=12)
    mean_tpr = 0.0
    mean_fpr = np.linspace(0., 1., 30)

    auc_all = []

    folds = StratifiedKFold(Y, n_folds=10,
                            shuffle=True,
                            random_state=np.random.randint(1, 100))

    # num_of_exp = 1
    for i, (train, test) in enumerate(folds):
        print "?%d?." % i
        # x_train, x_test, y_train, y_test = \
        #     train_test_split(X, Y,
        #                      test_size=ratio,
        #                      random_state=np.random.randint(1, 100))
        x_train, y_train = X[train], Y[train]
        x_test, y_test = X[test], Y[test]
        clf.fit(x_train, y_train)
        y_pred = clf.predict_proba(x_test)[:, 1]
        fpr, tpr, _ = metrics.roc_curve(y_test, y_pred)
        mean_tpr += interp(mean_fpr, fpr, tpr)
        mean_tpr[0] = 0.0
        auc_all.append(metrics.roc_auc_score(y_test, y_pred))

    mean_tpr /= len(folds)

    auc_array = np.array(auc_all)
    auc = auc_array.mean()
    auc_std = auc_array.std()
    mean_tpr[-1] = 1.0
    return mean_tpr, auc, auc_std