pip install fasttext

import jieba
import pandas as pd
import random

cate_dic = {'technology':1, 'car':2, 'entertainment':3, 'military':4, 'sports':5}

df_technology = pd.read_csv("./origin_data/technology_news.csv", encoding='utf-8')
df_technology = df_technology.dropna()

df_car = pd.read_csv("./origin_data/car_news.csv", encoding='utf-8')
df_car = df_car.dropna()

df_entertainment = pd.read_csv("./origin_data/entertainment_news.csv", encoding='utf-8')
df_entertainment = df_entertainment.dropna()

df_military = pd.read_csv("./origin_data/military_news.csv", encoding='utf-8')
df_military = df_military.dropna()

df_sports = pd.read_csv("./origin_data/sports_news.csv", encoding='utf-8')
df_sports = df_sports.dropna()

technology = df_technology.content.values.tolist()[1000:21000]
car = df_car.content.values.tolist()[1000:21000]
entertainment = df_entertainment.content.values.tolist()[:20000]
military = df_military.content.values.tolist()[:20000]
sports = df_sports.content.values.tolist()[:20000]

stopwords=pd.read_csv("origin_data/stopwords.txt",index_col=False,quoting=3,sep="\t",names=['stopword'], encoding='utf-8')
stopwords=stopwords['stopword'].values

#分词去停用词，并整理为fasttext要求的文本格式
def preprocess_text(content_lines, sentences, category):
    for line in content_lines:
        try:
            segs=jieba.lcut(line)
            segs = list(filter(lambda x:len(x)>1, segs))
            segs = list(filter(lambda x:x not in stopwords, segs))
            sentences.append("__label__"+str(category)+"\t"+" ".join(segs))
        except Exception as e:
            print(line)
            continue

#生成训练数据
sentences = []

preprocess_text(technology, sentences, cate_dic['technology'])
preprocess_text(car, sentences, cate_dic['car'])
preprocess_text(entertainment, sentences, cate_dic['entertainment'])
preprocess_text(military, sentences, cate_dic['military'])
preprocess_text(sports, sentences, cate_dic['sports'])

#数据打乱
random.shuffle(sentences)

# 写入数据-fasttext格式
def generate_model_data(sentences):
    train_num=int(len(sentences)*0.8)
    train_set=sentences[0:train_num]
    test_set=sentences[train_num:-1]
    print("writing data to fasttext format...")
    with open('./data/train_data.txt', 'w') as out:
        for sentence in train_set:
            out.write(sentence+"\n")
        print("done!")
    with open('./data/test_data.txt','w') as f:
        for sentence in test_set:
            f.write(sentence+'\n')
        print('done!')


generate_model_data(sentences)

import fasttext

classifier = fasttext.train_supervised('./data/train_data.txt',label='__label__', wordNgrams=2,epoch=20,lr=0.1,dim=100)

#参数说明
'''
train_supervised(input, lr=0.1, dim=100,
                   ws=5, epoch=5, minCount=1,
                   minCountLabel=0, minn=0,
                   maxn=0, neg=5, wordNgrams=1,
                   loss="softmax", bucket=2000000,
                   thread=12, lrUpdateRate=100,
                   t=1e-4, label="__label__",
                   verbose=2, pretrainedVectors="")
'''

  """
  训练一个监督模型, 返回一个模型对象

  @param input: 训练数据文件路径
  @param lr:              学习率
  @param dim:             向量维度
  @param ws:              cbow模型时使用
  @param epoch:           次数
  @param minCount:        词频阈值, 小于该值在初始化时会过滤掉
  @param minCountLabel:   类别阈值，类别小于该值初始化时会过滤掉
  @param minn:            构造subword时最小char个数
  @param maxn:            构造subword时最大char个数
  @param neg:             负采样
  @param wordNgrams:      n-gram个数
  @param loss:            损失函数类型, softmax, ns: 负采样, hs: 分层softmax
  @param bucket:          词扩充大小, [A, B]: A语料中包含的词向量, B不在语料中的词向量
  @param thread:          线程个数, 每个线程处理输入数据的一段, 0号线程负责loss输出
  @param lrUpdateRate:    学习率更新
  @param t:               负采样阈值
  @param label:           类别前缀
  @param verbose:         ??
  @param pretrainedVectors: 预训练的词向量文件路径, 如果word出现在文件夹中初始化不再随机
  @return model object

  """

classifier.save_model('./model/fasttext.bin')

’‘’
@return [样本个数, 准确率, 召回率]
‘’‘
train_result=classifier.test('./data/train_data.txt')
print('train_precision:', train_result[1])
print('train_recall:', train_result[2])
print('Number of train examples:', train_result[0])
test_result=classifier.test('./data/test_data.txt')
print('test_precision:', test_result[1])
print('test_recall:', test_result[2])
print('Number of test examples:', test_result[0])  


打印结果：
train_precision: 0.9906387350876191
train_recall: 0.9906387350876191
Number of train examples: 70076
test_precision: 0.9036990524032423
test_recall: 0.9036990524032423
Number of test examples: 17518

label_to_cate = {1:'technology', 2:'car', 3:'entertainment', 4:'military', 5:'sports'}

texts = '中新网 日电 2018 预赛 亚洲区 强赛 中国队 韩国队 较量 比赛 上半场 分钟 主场 作战 中国队 率先 打破 场上 僵局 利用 角球 机会 大宝 前点 攻门 得手 中国队 领先'
# texts = '这 是 中国 第 一 次 军舰 演习'
labels = classifier.predict(texts)
print(labels)
print(label_to_cate[int(labels[0][0].strip('__label__'))])


打印结果：
(('__label__5',), array([0.9999727]))
sports

model = fasttext.load_model(path)