在 pandas 的 split-apply-combine 期间省略的列答案

【问题标题】：A column that's omitted during split-apply-combie in pandas在 pandas 的 split-apply-combine 期间省略的列
【发布时间】：2016-04-11 22:59:41
【问题描述】：

我正在执行拆分-应用-组合来查找每个成员的总数量。我需要的数据框应该有 14 列：MemberID, DSFS_0_1, DSFS_1_2, DSFS_2_3, DSFS_3_4, DSFS_4_5, DSFS_5_6, DSFS_6_7, DSFS_7_8, DSFS_8_9, DSFS_9_10, DSFS_10_11, DSFS_11_12, DrugCount。但是，我没有得到第 14 个 (DrugCount)，知道为什么吗？变量 joined 输出全部 14，但我在其中进行聚合的函数 joined_grouped_add 仅返回 13。

import pandas as pd
import numpy as np
import matplotlib.pyplot as plt
import sys
from sklearn.cross_validation import train_test_split
from sklearn import linear_model

# this function takes the drugcount dataframe as input and output a tuple of 3 data frames: DrugCount_Y1,DrugCount_Y2,DrugCount_Y3
def process_DrugCount(drugcount):
    dc = pd.read_csv("DrugCount.csv")
    sub_map = {'1' : 1, '2':2, '3':3, '4':4, '5':5, '6':6, '7+' : 7}
    dc['DrugCount'] = dc.DrugCount.map(sub_map)
    dc['DrugCount'] = dc.DrugCount.astype(int)
    dc_grouped = dc.groupby(dc.Year, as_index=False)
    DrugCount_Y1 = dc_grouped.get_group('Y1')
    DrugCount_Y2 = dc_grouped.get_group('Y2')
    DrugCount_Y3 = dc_grouped.get_group('Y3')
    DrugCount_Y1.drop('Year', axis=1, inplace=True)
    DrugCount_Y2.drop('Year', axis=1, inplace=True)
    DrugCount_Y3.drop('Year', axis=1, inplace=True)
    return (DrugCount_Y1,DrugCount_Y2,DrugCount_Y3)

# this function converts strings such as "1- 2 month" to "1_2"
def replaceMonth(string):
    replace_map = {'0- 1 month' : "0_1", "1- 2 months": "1_2", "2- 3 months": "2_3", "3- 4 months": '3_4', "4- 5 months": "4_5", "5- 6 months": "5_6", "6- 7 months": "6_7", \
                   "7- 8 months" : "7_8", "8- 9 months": "8_9", "9-10 months": "9_10", "10-11 months": "10_11", "11-12 months": "11_12"}
    a_new_string = string.map(replace_map)
    return a_new_string

# this function processes a yearly drug count data
def process_yearly_DrugCount(aframe):
    processed_frame = None
    aframe.drop("Year", axis = 1, inplace = True)
    reformed = aframe[['DSFS']].apply(replaceMonth)
    gd = pd.get_dummies(reformed)
    joined =  pd.concat([aframe, gd], axis = 1)
    joined.drop("DSFS", axis = 1, inplace = True)
    joined_grouped = joined.groupby("MemberID", as_index = False)
    joined_grouped_agg = joined_grouped.agg(np.sum)
    print joined_grouped_agg
    return processed_frame
def main():
    pd.options.mode.chained_assignment = None 
    daysinhospital = pd.read_csv('DaysInHospital_Y2.csv')
    drugcount = pd.read_csv('DrugCount.csv')
    process_DrugCount(drugcount)
    process_yearly_DrugCount(drugcount)
    replaceMonth(drugcount['DSFS'])

if __name__ == '__main__':
    main()

【问题讨论】：

调用函数的行在哪里？
太，这里有太多的帮助。我建议打破每个部分并添加打印语句以查看内容以查看删除列的位置。否则，设置reproducible example。
我一路打印输出并将其分解。这就是我知道一切都很好的方式，直到我进行聚合joined_grouped_agg

标签： python pandas dataframe aggregate-functions split-apply-combine

【解决方案1】：

简单地说，直接从 csv 中提取的 DrugCount 不会作为数字字段 (int/float) 读取。否则它将保留在.agg(np.sum) 处理中。在聚合之前检查 dtype 并查看它是否是 object 类型（即字符串列）：

print joined['DrugCount'].dtype

事实上，在您的process_DrugCount() 函数中，您使用 astype 将 DrugCount 列显式转换为整数，但在 process_yearly_DrugCount() 函数中没有这样做。在后一个函数中运行同一行，DrugCount 应保留在聚合和处理中：

aframe['DrugCount'] = aframe['DrugCount'].astype(int)

或者更好的是，在main() 中避免在后面的函数中进行两次转换：

drugcount['DrugCount'] = drugcount['DrugCount'].astype(int)

另外，请注意，read_csv() 允许使用其 dtype 参数显式指定列类型：

drugcount = pd.read_csv('DrugCount.csv', dtype={'DrugCount': np.int64})

【讨论】：