fasttext初步使用

转载自：
http://blog.csdn.net/lxg0807/article/details/52960072#comments
训练数据和测试数据来自网盘：
https://pan.baidu.com/s/1jH7wyOY https://pan.baidu.com/s/1slGlPgx
训练以上数据

# _*_coding:utf-8 _*_
import logging
logging.basicConfig(format='%(asctime)s : %(levelname)s : %(message)s', level=logging.INFO)
import fasttext
#训练模型
classifier = fasttext.supervised("news_fasttext_train.txt","news_fasttext.model",label_prefix="__label__")

进行测试：

# -*- coding:utf-8 -*-

import logging
logging.basicConfig(format='%(asctime)s : %(levelname)s : %(message)s', level=logging.INFO)
import fasttext

#load训练好的模型
classifier = fasttext.load_model('news_fasttext.model.bin', label_prefix='__label__')
result = classifier.test("news_fasttext_test.txt")
print result.precision
print result.recall

注意每次训练的模型都有不同，所以测试的结果大概是0.87~0.92左右

进行最终评价：

# -*- coding:utf-8 -*-

import logging
logging.basicConfig(format='%(asctime)s : %(levelname)s : %(message)s', level=logging.INFO)
import fasttext

#load训练好的模型
classifier = fasttext.load_model('news_fasttext.model.bin', label_prefix='__label__')
result = classifier.test("news_fasttext_test.txt")
print result.precision
print result.recall
labels_right = []
texts = []
with open("news_fasttext_test.txt") as fr:
lines = fr.readlines()
for line in lines:
labels_right.append(line.split("\t")[1].rstrip().replace("__label__",""))
texts.append(line.split("\t")[0].decode("utf-8"))
#     print labels
#     print texts
#     break
labels_predict = [e[0] for e in classifier.predict(texts)] #预测输出结果为二维形式
# print labels_predict

text_labels = list(set(labels_right))
text_predict_labels = list(set(labels_predict))
print text_predict_labels
print text_labels

A = dict.fromkeys(text_labels,0)  #预测正确的各个类的数目
B = dict.fromkeys(text_labels,0)   #测试数据集中各个类的数目
C = dict.fromkeys(text_predict_labels,0) #预测结果中各个类的数目
for i in range(0,len(labels_right)):
B[labels_right[i]] += 1
C[labels_predict[i]] += 1
if labels_right[i] == labels_predict[i]:
A[labels_right[i]] += 1

print A
print B
print C
#计算准确率，召回率，F值
for key in B:
p = float(A[key]) / float(B[key])
r = float(A[key]) / float(C[key])
f = p * r * 2 / (p + r)
print "%s:\tp:%f\t%fr:\t%f" % (key,p,r,f)

之所以搞这么一出，是因为fasttext提供的p值（准确率）和r值（召回率）只是针对所有结果的，而不是针对各个类别分别进行计算p值（准确率）和r值（召回率）的，所以该作者自己写了计算方法。

输出结果：

[u'affairs', u'fashion', u'lottery', u'house', u'sports', u'game', u'economic', u'ent', u'edu', u'home', u'stock', u'constellation', u'science']
['affairs', 'fashion', 'house', 'sports', 'game', 'economic', 'ent', 'edu', 'home', 'stock', 'science']
{'science': 8921, 'affairs': 8544, 'fashion': 2148, 'house': 9572, 'sports': 9814, 'game': 9389, 'economic': 9492, 'ent': 9660, 'edu': 9671, 'home': 8027, 'stock': 8525}
{'science': 10000, 'affairs': 10000, 'fashion': 3369, 'house': 10000, 'sports': 10000, 'game': 10000, 'economic': 10000, 'ent': 10000, 'edu': 10000, 'home': 10000, 'stock': 10000}
{u'science': 10311, u'affairs': 8953, u'fashion': 2176, u'lottery': 28, u'house': 10502, u'sports': 10288, u'game': 10182, u'economic': 11087, u'ent': 10940, u'edu': 10991, u'home': 8171, u'constellation': 466, u'stock': 9274}
science:	p:0.892100	0.865193r:	0.878440
affairs:	p:0.854400	0.954317r:	0.901599
fashion:	p:0.637578	0.987132r:	0.774752
house:	p:0.957200	0.911445r:	0.933763
sports:	p:0.981400	0.953927r:	0.967468
game:	p:0.938900	0.922117r:	0.930433
economic:	p:0.949200	0.856138r:	0.900270
ent:	p:0.966000	0.882998r:	0.922636
edu:	p:0.967100	0.879902r:	0.921443
home:	p:0.802700	0.982377r:	0.883496
stock:	p:0.852500	0.919237r:	0.884611