xgboost实例代码

 1 # -*- coding: utf-8 -*-
 2 import xgboost as xgb
 3 import csv
 4 import jieba
 5 jieba.load_userdict(‘wordDict.txt‘)
 6 import numpy as np
 7 from sklearn.feature_extraction.text import CountVectorizer
 8 from sklearn.feature_extraction.text import TfidfTransformer
 9  
10  
11 # 读取训练集
12 def readtrain():
13     with open(‘Train.csv‘, ‘rb‘) as csvfile:
14         reader = csv.reader(csvfile)
15         column1 = [row for row in reader]
16     content_train = [i[1] for i in column1[1:]] # 第一列为文本内容，并去除列名
17     opinion_train = [i[2] for i in column1[1:]] # 第二列为类别，并去除列名
18     print ‘训练集有 %s 条句子‘ % len(content_train)
19     train = [content_train, opinion_train]
20     return train
21  
22  
23 # 将utf8的列表转换成unicode
24 def changeListCode(b):
25     a = []
26     for i in b:
27         a.append(i.decode(‘utf8‘))
28     return a
29  
30  
31 # 对列表进行分词并用空格连接
32 def segmentWord(cont):
33     c = []
34     for i in cont:
35         a = list(jieba.cut(i))
36         b = " ".join(a)
37         c.append(b)
38     return c
39  
40  
41 # 类别用数字表示：pos:2,neu:1,neg:0
42 def transLabel(labels):
43     for i in range(len(labels)):
44         if labels[i] == ‘pos‘:
45             labels[i] = 2
46         elif labels[i] == ‘neu‘:
47             labels[i] = 1
48         elif labels[i] == ‘neg‘:
49             labels[i] = 0
50         else: print "label无效：",labels[i]
51     return labels
52  
53  
54 train = readtrain()
55 content = segmentWord(train[0])
56 opinion = transLabel(train[1])  # 需要用数字表示类别
57 opinion = np.array(opinion)     # 需要numpy格式
58  
59  
60 train_content = content[:7000]
61 train_opinion = opinion[:7000]
62 test_content = content[7000:]
63 test_opinion = opinion[7000:]
64  
65  
66 vectorizer = CountVectorizer()
67 tfidftransformer = TfidfTransformer()
68 tfidf = tfidftransformer.fit_transform(vectorizer.fit_transform(train_content))
69 weight = tfidf.toarray()
70 print tfidf.shape
71 test_tfidf = tfidftransformer.transform(vectorizer.transform(test_content))
72 test_weight = test_tfidf.toarray()
73 print test_weight.shape
74  
75  
76 dtrain = xgb.DMatrix(weight, label=train_opinion)
77 dtest = xgb.DMatrix(test_weight, label=test_opinion)  # label可以不要，此处需要是为了测试效果
78 param = {‘max_depth‘:6, ‘eta‘:0.5, ‘eval_metric‘:‘merror‘, ‘silent‘:1, ‘objective‘:‘multi:softmax‘, ‘num_class‘:3}  # 参数
79 evallist  = [(dtrain,‘train‘), (dtest,‘test‘)]  # 这步可以不要，用于测试效果
80 num_round = 50  # 循环次数
81 bst = xgb.train(param, dtrain, num_round, evallist)
82 preds = bst.predict(dtest)