Python 3.x Scikit最近邻的TFIDF点列表

我可以为一个TFIDF运行kneighbors，但不能为它们的列表运行kneighbors

在详细介绍之前，我应该提到我这样做的原因，因为为每个数据点运行kneighbors需要很长时间，我想给kneighbors一个点列表将在内部进行优化

根据NN文档：

它说我可以查询多个点：

>>>X = [[0., 1., 0.], [1., 0., 1.]]
>>>neigh.kneighbors(X, return_distance=False) 
>>>array([[1],
   [2]]...)

我也试着这么做。 我可以分别为每个点运行kneighbors：

from sklearn.neighbors import NearestNeighbors
from sklearn.feature_extraction.text import CountVectorizer, TfidfTransformer

samples = ["This is a test","a very good test","some more text"]
count_vect = CountVectorizer()
X_train_counts = count_vect.fit_transform(samples)
tfidf_transformer = TfidfTransformer()
X_train_tfidf = tfidf_transformer.fit_transform(X_train_counts)
neigh = NearestNeighbors(n_neighbors=1, n_jobs=-1) 
neigh.fit(X_train_tfidf)

ll=[]
test=["Test if this works","Zoom zoom"]
for k in test:
    predict = count_vect.transform([k])
    X_tfidf2 = tfidf_transformer.transform(predict)
    ll.append(X_tfidf2)
    res = neigh.kneighbors(X_tfidf2, return_distance=False)
#res = neigh.kneighbors(ll, return_distance=False)

当我将所有TFIDF稀疏矩阵添加到列表并尝试时，我得到一个错误。取消注释最后一行以获取错误

错误： ValueError:使用序列设置数组元素（在res=neigh.kneighbors…行上）

尝试：

from scipy import sparse

from sklearn.neighbors import NearestNeighbors
from sklearn.feature_extraction.text import CountVectorizer, TfidfTransformer

samples = ["This is a test","a very good test","some more text"]
count_vect = CountVectorizer()
X_train_counts = count_vect.fit_transform(samples)
tfidf_transformer = TfidfTransformer()
X_train_tfidf = tfidf_transformer.fit_transform(X_train_counts)
neigh = NearestNeighbors(n_neighbors=1, n_jobs=-1) 
neigh.fit(X_train_tfidf)

ll=[]
test=["Test if this works","Zoom zoom"]
for k in test:
    predict = count_vect.transform([k])
    X_tfidf2 = tfidf_transformer.transform(predict)
    ll.append(X_tfidf2)

ll = sparse.vstack((ll))
res = neigh.kneighbors(ll, return_distance=False)

没有循环： 从scipy导入稀疏

from sklearn.neighbors import NearestNeighbors
from sklearn.feature_extraction.text import CountVectorizer, TfidfTransformer

samples = ["This is a test","a very good test","some more text"]
count_vect = CountVectorizer()
X_train_counts = count_vect.fit_transform(samples)
tfidf_transformer = TfidfTransformer()
X_train_tfidf = tfidf_transformer.fit_transform(X_train_counts)
neigh = NearestNeighbors(n_neighbors=1, n_jobs=-1) 
neigh.fit(X_train_tfidf)

test=["Test if this works","Zoom zoom"]
X_test_counts = count_vect.transform(test)

X_test_tfidf = tfidf_transformer.transform(X_test_counts)

res = neigh.kneighbors(X_test_tfidf, return_distance=False)

---

你能发布一些实际运行的示例代码吗？我用演示数据更新了文章，使其运行。当我尝试这样做时，我得到：ValueError:X和Y矩阵的不兼容维度。X.shape[1]==16而Y.shape[1]==8似乎有效：）测试结果以确保它们是相同的，并且将接受答案@兄弟姐妹你不需要使用循环顺便说一句。见更新的答案。真的！再次感谢。顺便说一句，由于所有邻居都是同时计算的，所以它的运行速度要快得多！