最近在用python做數據挖掘，在聚類的時候遇到了一個非常惡心的問題。話不多說，直接上代碼：
1 from sklearn.cluster import KMeans 2 from sklearn.decomposition import PCA 3 import matplotlib.pyplot as plt 4 #kmeans算法 5 df1=df23 6 kmeans = KMeans(n_clusters=5, random_state=10).fit(df1) 7 #貼上每個樣本對應的簇類別標簽 8 df1['level']=kmeans.labels_ 9 #df1.to_csv('new_df.csv') 10 11 df2=df1.groupby('level',as_index=False)['level'].agg({'num': np.size}) 12 print(df2.head()) 13 14 #將用於聚類的數據的特征的維度降至2維 15 pca = PCA(n_components=2) 16 new_pca = pd.DataFrame(pca.fit_transform(df1)) 17 print(new_pca.head()) 18 19 #可視化 20 d = new_pca[df1['level'] == 0] 21 plt.plot(d[0], d[1], 'gv') 22 d = new_pca[df1['level'] == 1] 23 plt.plot(d[0], d[1], 'ko') 24 d = new_pca[df1['level'] == 2] 25 plt.plot(d[0], d[1], 'b*') 26 d = new_pca[df1['level'] == 3] 27 plt.plot(d[0], d[1], 'y+') 28 d = new_pca[df1['level'] == 4] 29 plt.plot(d[0], d[1], 'c.') 30 31 plt.title('the result of polymerization') 32 plt.show()
錯誤如下：
網上找了好久都沒找到解決方法，明明之前成功過的。於是我查看了df23數據，發現它是這樣的：
與之前成功的dataframe的唯一差別就是索引！！！重要的事情說三遍！！！索引！！！索引！！！於是乎，我去找怎么重置索引的方法，見代碼：
1 df24=df23[["forks_count","has_issues","has_wiki","open_issues_count","stargazers_count","watchers_count","created_pushed_time","created_updated_time"]] 2 df24=df24.reset_index() 3 df24=df24[["forks_count","has_issues","has_wiki","open_issues_count","stargazers_count","watchers_count","created_pushed_time","created_updated_time"]]
然后聚類就成功了。。。心累。。。。