时间戳

首页 > 解决方案 > skmultiLearn 分类器预测总是返回 0

python - skmultiLearn 分类器预测总是返回 0

问题描述

我对 skmultiLearn 很陌生，现在我将它用于“中文”文档的多标签分类。训练数据集很小（比如 200 句话），我一共设置了 6 个类。即使我使用句子 IN 训练数据集，我也只能得到 [0,0,0,0,0,0] 作为预测结果，我能得到一些帮助吗？谢谢！

我的代码：

# Import BinaryRelevance from skmultilearn
from skmultilearn.problem_transform import BinaryRelevance

from sklearn.feature_extraction.text import CountVectorizer, TfidfVectorizer
from sklearn.feature_extraction.text import TfidfTransformer
from sklearn.svm import SVC
from scipy import sparse 
import jieba
import codecs
import numpy as np

from warnings import simplefilter
simplefilter(action='ignore', category=FutureWarning)

Q_list = []
L_list = []

# Read Sentence file
with codecs.open('multi-label-Q.txt',encoding='utf-8') as infile:
    for line in infile:
        Q_list.append(line[:-2])
infile.close()

# Read Label file
with open('multi-label-L.txt') as infile:
    for line in infile:
        tmp_l = line[:-1].split(',')
        L_list.append(tmp_l)
infile.close()

L_list = np.array(L_list)

L_Question_list = []

# Preprocess for Chinese sentences
for line in Q_list:
    seg_list = jieba.lcut(line, cut_all=False)
    q_addSpace = ''
    for w in seg_list:
        q_addSpace = q_addSpace + w + ' '
    L_Question_list.append(q_addSpace[:-1])


cv = CountVectorizer()
cv_fit=cv.fit_transform(L_Question_list)

transformer = TfidfTransformer()
tfidf = transformer.fit_transform(cv_fit)

M = sparse.lil_matrix((len(L_list),6), dtype=int)
for i,row in enumerate(L_list):
    count = 0
    for col in row:
        M[i, count] = col
        count += 1

# Setup the classifier
clf = BinaryRelevance(classifier=SVC())



# Train
clf.fit(tfidf, M)

# A sentence in train dataset
x_test = '偏头痛多发于什么年龄层？'
# Preprocess for Chinese sentence
seg_list = jieba.lcut(x_test, cut_all=False)
q_addSpace = ''
for w in seg_list:
    q_addSpace = q_addSpace + w + ' '
X_test = [q_addSpace]
cv_fit2=cv.transform(X_test)
tfidf2 = transformer.transform(cv_fit2)


# Predict
pred = clf.predict(tfidf2)
print(pred.todense())

标签： pythonscikit-learnmultilabel-classification

解决方案

现在我明白了，原因是我有太多的单标签数据。

我使用了一些高价值的数据集并得到了正确的结果。

所以，答案是：完善数据集。

推荐阅读