对每块训练集的前99000数据训练，后1000数据集进行测试 

 1 from sklearn.neighbors import KNeighborsClassifier
 2 from sklearn.externals import joblib
 3 onehot = OneHotEncoder()
 4 for b in range(1,115):
 5     addata = pd.read_csv("adFeature.csv")
 6     userdata = pd.read_csv("userFeature_%d_part_bat.csv"%(b*100000))
 7     traindata = pd.read_csv("train.csv")
 8     data = pd.merge(userdata,traindata)
 9     data = pd.merge(data,addata)
10     #data.to_csv("111.csv")
11     #列出所有特征值，遍历,uid是用户唯一标识，不应该算作特征
12 
13     userfeature = ["age","carrier","consumptionAbility","ct","education","gender","house","interest1","interest2","interest3","interest4","interest5","kw1","kw2","kw3","marriageStatus","os","topic1","topic2","topic3","LBS","appIdAction","appIdInstall","campaignId","creativeId","creativeSize","adCategoryId","advertiserId","productId","productType"]
14     #for index in data[feature] :
15     userdata = []
16     for index in range(len(data["uid"])):
17         feature_li = []
18         for feature in userfeature:
19             # a = data[feature]
20             # print(a[0],type(a[index]),isinstance(a[0],(numpy.int64)))
21 
22             if isinstance(data[feature][index],numpy.int64):
23                 feature_li.append(int(data[feature][index]))
24             elif isinstance(data[feature][index],numpy.float64):
25                 feature_li.append(0)#缺失值用0填充，这是不合理的，有待改进
26             elif isinstance(data[feature][index], numpy.float):
27                 feature_li.append(0)
28             else :
29                 trans = data[feature][index].strip().split(" ")
30                 trans = map(int,trans)
31                 trans = sorted(trans)
32                 #print(trans)
33                 s = 0
34                 for num in trans :
35                     s += num
36                 feature_li.append(s)
37         # print(feature_li)
38         userdata.append(feature_li)
39     userdata = numpy.array(userdata)
40     onehot.fit(userdata)
41     print("--------------------------------------------------------------------")
42     Y = numpy.array(data["label"])
43     print("--------------------------------------------------------------------")
44     X = onehot.transform(userdata)
45     print(X)
46     print(numpy.shape(X))
47     knn = KNeighborsClassifier()
48     model = knn.fit(X[:99000],Y[:99000])
49     joblib.dump(model,"%d.model"%(b))
50     result = model.predict(X[-900:])
51     print("-----------------------------------------------------------------------")
52     #print(Y[-900:])
53     print("--------------------------------------------------------------------------")
54     s = 0
55     for i in range(len(result)):
56         if result[i] == Y[-900:][i] :
57             s +=1
58     a = s/len(result)
59     f = open("result.txt","a",encoding="utf-8")
60     f.write(str(a))
61     print(a)

 

posted @ 2018-04-25 11:23  我想休息  阅读(331)  评论(0)    收藏  举报