我有一个带有费用清单的文件:
示例:
date item out in
12/01/2017 PAGO FIBERTEL 668.5 0.0
12/01/2017 PAGO GAS NATURAL 2.32 0.0
10/02/2017 EXTRACCION TARJETA 1200.0 0.0
10/02/2017 CPA. STARBUCKS R. PE9A 105.0 0.0
10/02/2017 CPA. STARBUCKS R. PE9A 125.0 0.0
11/03/2017 EXTRACCION TARJETA 1200.0 0.0
11/03/2017 SALES 0.0 10000.0
我想制作一个绘图,在其中可以查看每年一年中某些项目的演变情况。例如,我将使用“ startbucks”作为关键字过滤“ item”列,我将计算每月汇总,并显示如下信息:
Dec Jan Mar
Starbucks 0 0 230
我从json文件中提取了一系列关键字,这些关键字将用于产生每一行。但是,我不能只用一个。我已经尝试了多种形式的groupby(使用grouper和不使用grouper),但是我认为我没有。这是我目前获得的代码:
import pandas as pd
import matplotlib.pyplot as plt
import sys
import json
class Banca():
def __init__(self, name, csv_path, json_path):
self.name= name
self.df = pd.read_csv(csv_path)
with open(json_path) as j:
self.json = json.load(j)
def prepare(self):
#Add header
headers = ['fecha','concepto','in','out',"x"]
self.df.columns = headers
#fix data
self.df.fecha = pd.to_datetime(self.df.fecha)
#Here i'm testing, this doesnt work
g1=self.df.groupby(pd.Grouper(key='fecha', freq='M')['in'].sum())
print(g1.describe().to_string())
print(g1.head())
#g1.plot(y='out', style='.-', figsize=(15,4))
#plt.show()
#filter data
# some filter
def grafica(self):
#plot data
self.df.plot(x='fecha', y='out',style='.-', figsize=(15,4))
plt.show()
def test_df(self):
print(self.df.describe(include='all'))
def test_json(self):
for x,y in self.json.items():
print(x,y)
icbc = Banca("ICBC", sys.argv[1], sys.argv[2])
icbc.test_df()
icbc.prepare()
#icbc.grafica()
#icbc.test_json()
我正在编写此代码,作为学习熊猫操作数据的练习。我已经学到了很多关节,但是我已经在这里停留了一段时间。我在想也许我不应该为此使用groupby,而是其他。无论如何,我感谢您的帮助。
答案 0 :(得分:1)
使用:
#convert column to datetimes if necessary
df['fecha'] = pd.to_datetime(df['fecha'], format='%d/%m/%Y')
print(df)
fecha concepto in out
0 2017-01-12 PAGO FIBERTEL 668.50 0.0
1 2017-01-12 PAGO GAS NATURAL 2.32 0.0
2 2017-02-10 EXTRACCION TARJETA 1200.00 0.0
3 2017-02-10 CPA. STARBUCKS R. PE9A 105.00 0.0
4 2017-02-10 CPA. STARBUCKS R. PE9A 125.00 0.0
5 2017-03-11 EXTRACCION TARJETA 1200.00 0.0
6 2017-03-11 SALES 0.00 10000.0
import re
#create DatetimeIndex
df = df.set_index('fecha')
#list of values
L = ['starbuck','pago']
all_s = []
for x in L:
#filter by substrings, select column in
s = df.loc[df['concepto'].str.contains(x, flags=re.I), 'in']
#aggregate by months and sum
s = s.groupby(pd.Grouper(freq='M')).sum()
#change format of index by `MM-YYYY`
s.index = s.index.strftime('%b-%Y')
all_s.append(s.rename(x))
#join all Series together and transpose
df = pd.concat(all_s, axis=1).T
print (df)
Feb-2017 Jan-2017
starbuck 230.0 NaN
pago NaN 670.82
编辑:
对于绘图,最好绘制DatetimeIndex
和按关键字划分的列,还可以对月份的开始按MS
分组,如果要添加由0
填充的缺失月份,则可以添加asfreq
。 :
df['fecha'] = pd.to_datetime(df['fecha'], format='%d/%m/%Y')
print(df)
fecha concepto in out
0 2017-01-12 PAGO FIBERTEL 668.50 0.0
1 2017-01-12 PAGO GAS NATURAL 2.32 0.0
2 2017-02-10 EXTRACCION TARJETA 1200.00 0.0
3 2017-02-10 CPA. STARBUCKS R. PE9A 105.00 0.0
4 2017-02-10 CPA. STARBUCKS R. PE9A 125.00 0.0
5 2017-03-11 EXTRACCION TARJETA 1200.00 0.0
6 2017-05-11 SALES 20.00 10000.0 <-changed last month
import re
df = df.set_index('fecha')
L = ['starbuck','pago', 'sales']
all_s = []
for x in L:
s = df.loc[df['concepto'].str.contains(x, flags=re.I), 'in']
s = s.groupby(pd.Grouper(freq='MS')).sum()
all_s.append(s.rename(x))
df = pd.concat(all_s, axis=1).fillna(0).asfreq('MS', fill_value=0)
print (df)
starbuck pago sales
fecha
2017-01-01 0.0 670.82 0.0
2017-02-01 230.0 0.00 0.0
2017-03-01 0.0 0.00 0.0
2017-04-01 0.0 0.00 0.0
2017-05-01 0.0 0.00 20.0
df.plot(style='.-', figsize=(15,4))