Question

我有一个数据集，我已经运行了K-means算法（scikit-learn），我想在每个集群上构建一个决策树。我可以从群集中恢复值，但不能恢复＆＃34;类＆＃34;值（我做监督学习，每个元素可以属于两个类中的一个，我需要与数据相关的值来构建我的树）

Ex：未经过滤的数据集：

[val1 val2 class]
X_train=[val1 val2]
y_train=[class]

聚类代码如下：

X = clusterDF[clusterDF.columns[clusterDF.columns.str.contains('\'AB\'')]]
y = clusterDF['Class']
(X_train, X_test, y_train, y_test) = train_test_split(X, y,
        test_size=0.30)

kmeans = KMeans(n_clusters=3, n_init=5, max_iter=3000, random_state=1)
kmeans.fit(X_train, y_train)
y_pred = kmeans.predict(X_test)

这是我（令人难以置信的笨拙！）代码，用于提取构建树的值。问题是Y值;它们与X值不一致

cl={i: np.where(kmeans.labels_ == i)[0] for i in range(kmeans.n_clusters)}
for j in range(0,len(k_means_labels_unique)):
    Xc=None
    Y=None
    #for i in range(0,len(k_means_labels_unique)):
    indexes = cl.get(j,0)
    for i, row in X.iterrows():
        if i in indexes:
            if Xc is not None:
                Xc = np.vstack([Xc, [row['first occurrence of \'AB\''],row['similarity to \'AB\'']]])
            else:
                Xc = np.array([row['first occurrence of \'AB\''],row['similarity to \'AB\'']])
            if Y is not None:
                Y = np.vstack([Y, y[i]])
            else:
                Y = np.array(y[i])
    Xc = pd.DataFrame(data=Xc, index=range(0, len(X)),
                     columns=['first occurrence of \'AB\'',
        'similarity to \'AB\''])  # 1st row as the column names


    Y = pd.DataFrame(data=Y, index=range(0, len(Y)),columns=['Class'])


        print("\n\t-----Classifier ", j + 1,"----")

        (X_train, X_test, y_train, y_test) = train_test_split(X, Y,
            test_size=0.30)

        classifier = DecisionTreeClassifier(criterion='entropy',max_depth = 2)
        classifier = getResults(
            X_train,
        y_train,
        X_test,
        y_test,
        classifier,
        filename='classif'+str(3 + i),
        )

采用群集数据制作决策树的任何想法（或更有效的方法）？

Answer 1

没有阅读所有代码，但我的猜测是将索引向量传递到train_test_split函数可以帮助您跟踪样本。

X = clusterDF[clusterDF.columns[clusterDF.columns.str.contains('\'AB\'')]]
y = clusterDF['Class']
indices = clusterDF.index
X_train, X_test, y_train, y_test, indices_train, indices_test = train_test_split(X, y, indices)

在聚类之后从k-means集群获取值

1 个答案: