pandas kmeans如何使用分类属性

Question

我使用Cloudera 5.2 VM和pandas 0.18.0 我想将kmeans应用于我的数据帧。但我有str列。

我的数据框是

adClicksPerTime.head(n=5)
Out[50]: 
            timestamp   adCategory  userId  totalAdClicks
0 2016-05-26 15:00:00   automotive     355              1
1 2016-05-26 15:00:00     clothing    1027              1
2 2016-05-26 15:00:00    computers    1821              1
3 2016-05-26 15:00:00    computers    2139              1
4 2016-05-26 15:00:00  electronics     253              1

for col in adClicksPerTime:
     print(col)
     print(type(adClicksPerTime[col][1]))


timestamp
<class 'pandas.tslib.Timestamp'>
adCategory
<class 'str'>
userId
<class 'numpy.int64'>
totalAdClicks
<class 'numpy.int64'>

当我执行kmeans时，我收到错误

ValueError: could not convert string to float: 'automotive'

我尝试将我的字符串转换为分类类型，然后再分配数字代码

adClicksPerTime.adCategory = pd.Categorical.from_array(adClicksPerTime.adCategory)     

adClicksPerTime.head(n=5)
Out[54]: 
            timestamp   adCategory  userId  totalAdClicks
0 2016-05-26 15:00:00   automotive     355              1
1 2016-05-26 15:00:00     clothing    1027              1
2 2016-05-26 15:00:00    computers    1821              1
3 2016-05-26 15:00:00    computers    2139              1
4 2016-05-26 15:00:00  electronics     253              1

for col in adClicksPerTime:
     print(col)
     print(type(adClicksPerTime[col][1]))


timestamp
<class 'pandas.tslib.Timestamp'>
adCategory
<class 'str'>
userId
<class 'numpy.int64'>
totalAdClicks
<class 'numpy.int64'>

如何将kmeans应用于此str字段？

Answer 1

获取傻瓜会将类别更改为傻瓜。

dummies = pd.get_dummies(adClicksPerTime[adCategory])
del dummies['automotive']
print dummies.columns

然后将此DataFrame与adClicksPerTime dataFrame合并，最后应用Kmeans。

adClicksPerTime.info()会给你dtypes。