stats.py 文件源码

def n3_error_rate_nearest_neighbor_classifier(data):

    features = data.columns[:-1, ]
    mistakes = 0
    n = data.shape[0]

    for i in range(n):
        bad_df = data.index.isin([i])
        good_df = ~bad_df

        knn = KNeighborsClassifier( n_neighbors=1 )
        knn.fit(data.iloc[good_df].iloc[:, :-1], data.iloc[good_df].iloc[: ,-1])
        temp = np.array(data.iloc[i ,:-1]).reshape(1,-1)
        mistake = 1 if data.iloc[i, -1] != knn.predict(temp) else 0

        mistakes = mistakes + mistake

    n3 = (1.0 * mistakes) / n
    if n3 > 1:
        n3 = 1
    return n3