Scikit学习：为什么我的kmeans聚类散点图如此丑陋？

Question

def get_nouns (text):
    tagger = MeCab.Tagger()
    words = []

    for c in tagger.parse(text).splitlines()[:-1]:
        if len(c.split('\t')) < 2:
            continue
        surface, feature = c.split('\t')
        pos = feature.split(',')[0]
        if pos == '名詞':  # noun 
            words.append(surface)
    return ' '.join(words)


def bio():
    biolist =[]
    howmany = 10
    for giin in read:
        if len(giin["education"]) < 1:
            continue
        biolist.append(get_nouns(" ".join(giin["education"])))
    ######################################################
    nparray = np.array (biolist)
    cv = CountVectorizer()
    bags = cv.fit_transform(nparray)
    tfidf=TfidfTransformer(norm='l2', sublinear_tf=True).fit_transform(bags)
    km_model = KMeans(n_clusters=howmany, init='k-means++')
    km_model.fit_transform(tfidf)
    lsa2 = TruncatedSVD(2)
    compressed_text_list = lsa2.fit_transform(tfidf)
    compressed_center_list = lsa2.fit_transform(km_model.cluster_centers_)

    X = []
    Y = []
    X_cent = []
    Y_cent = []

    for x, y in compressed_text_list:
        X.append(x)
        Y.append(y)
    for x, y in compressed_center_list:
        X_cent.append(x)
        Y_cent.append(y)

    clus_list = []
    for i in range (howmany):
        clus_list.append([])
        for a in biolist:
            if km_model.labels_[biolist.index(a)] == i:
                clus_list[i].append(a)

    for a in clus_list:
        print (a)
        print ("   ")

    plt.scatter(X, Y,  c = km_model.labels_)
    plt.scatter(X_cent, Y_cent, c="r", marker = "+")

    plt.show()

我有一个代码，将人们的教育历史分为10组。我的散点图是这样的。

如您所见，散点图并没有真正按组排序，不同的颜色相互混合。如何更改此代码以使分组更精确