目录
2、使用python list、python dict、numpy.ndarray创建pandas.Series
3、将pandas.Series转化为pandas.DataFrame
4、将多个pandas.Series合并为一个pandas.DataFrame
6、 移除pandas.Series1中和pandas.Series2共同的部分
7、求pandas.Series1和pandas.Series2的交集、并集、差集
8、求pandas.Series分位数（最小值、1/4分位数、中位数、3/4分位数、最大值）
10、输出pandas.Series()中频数排第一二位的、其它的替换为other
11、将pandas.Series()均分为10个区间、每个值使用区间名称标记
12、将pandas.Series()转换为指定shape的pandas.DataFrame
13、取出pandas.Series()中满足条件数据的位置index
17、求真实和预测pd.Series之间的均方误差损失函数（MSE，mean squared error)
21、pd.Series中日期字符串转换为datetime格式
26、pd.Series1按pd.Series2分组并求均值
33、从一个csv 文件中每间隔50行取数据生成pandas.DataFrame
34、从一个csv 文件取数据生成pandas.DataFrame（新增加一分类列）
35、生成一个按规定步长平移的pandas.DataFrame
36、从一个csv 文件读取指定列生成pandas.DataFrame
37、输出DataFrame的行数、列数、数据类型、类型频数、Series转list
43、各自列均值、中值填充DataFrame中各自列缺省值（使用apply）
52、获取DataFrame某一列中第n大的值大于指定值得索引
54、 Series or DataFrame中使用分位数填充超限区域
60、DataFrame创建新列：每行为行号（按欧几里得距离而来）
62、DataFrame创建一列：包含每行中最小值与最大值比值
68、DataFrame按另外列分组、提取当前列中指定值（看下方例子，需求不好描述）
70、两个DataFrame使用类似SQL 中INNER JOIN拼接
73、DataFrame中新建两列：滞后列和提前列（看下方例子，需求BT）
1、pandas导入、查看版本
#pandas导入
import pandas as pd
#pandas输出版本信息
print(pd.__version__)
#pandas输出详细版本信息、 Python版本、相关程序包、操作系统等信息以json格式输出
print(pd.show_versions(as_json=True))
#pandas输出相关信息以默认格式输出
print(pd.show_versions())
2、使用python list、python dict、numpy.ndarray创建pandas.Series
import pandas as pd
import numpy as np
mylist = list('abcedfghijklmnopqrstuvwxyz')# python list
myarr = np.arange(26)#numpy.ndarray
mydict = dict(zip(mylist, myarr))#python dict
ser1 = pd.Series(mylist)
ser2 = pd.Series(myarr)
ser3 = pd.Series(mydict)
print(ser1.head(2))
print(ser2.head(2))
print(ser3.head(2))
3、将pandas.Series转化为pandas.DataFrame
需求：
解决：
mylist = list('abcedfghijklmnopqrstuvwxyz')
myarr = np.arange(26)
mydict = dict(zip(mylist, myarr))
ser = pd.Series(mydict)
#to_frame()结合reset_index()使用
df = ser.to_frame().reset_index()
print(df.head())
4、将多个pandas.Series合并为一个pandas.DataFrame
# 输入
import numpy as np
ser1 = pd.Series(list('abcedfghijklmnopqrstuvwxyz'))
ser2 = pd.Series(np.arange(26))
# 解决方法1
df = pd.concat([ser1, ser2], axis=1)
# 解决方法2
df = pd.DataFrame({'col1': ser1, 'col2': ser2})
print(df.head()) 5、修改pandas.Series index名称
6、 移除pandas.Series1中和pandas.Series2共同的部分
ser1 = pd.Series([1, 2, 3, 4, 5])
ser2 = pd.Series([4, 5, 6, 7, 8])
ser1[~ser1.isin(ser2)]
7、求pandas.Series1和pandas.Series2的交集、并集、差集
ser1 = pd.Series([1, 2, 3, 4, 5])
ser2 = pd.Series([4, 5, 6, 7, 8])
ser_u = pd.Series(np.union1d(ser1, ser2)) # union
ser_i = pd.Series(np.intersect1d(ser1, ser2)) # intersect
ser_s = ser_u[~ser_u.isin(ser_i)]# Subtraction
print("交集")
print(ser_i)
print("并集")
print(ser_u)
print("差集")
print(ser_s)
8、求pandas.Series分位数（最小值、1/4分位数、中位数、3/4分位数、最大值）
state = np.random.RandomState(100)
ser = pd.Series(state.normal(10, 5, 25))
np.percentile(ser, q=[0, 25, 50, 75, 100])#可是设置其它任何想输出的分位数 
9、求pandas.Series()频数
ser = pd.Series(np.take(list('abcdefgh'), np.random.randint(8, size=30)))
ser.value_counts()
10、输出pandas.Series()中频数排第一二位的、其它的替换为other
np.random.RandomState(100)
ser = pd.Series(np.random.randint(1, 5, [12]))
print("Top 2 Freq:", ser.value_counts())
ser[~ser.isin(ser.value_counts().index[:2])] = 'Other'
ser
11、将pandas.Series()均分为10个区间、每个值使用区间名称标记
np.random.seed(666) #让结果可重复
ser = pd.Series(np.random.random(20))
print(ser.head())
pd.qcut(ser,
q=[0, .10, .20, .3, .4, .5, .6, .7, .8, .9, 1],
labels=[
'1st', '2nd', '3rd', '4th', '5th', '6th', '7th', '8th', '9th',
'10th'
])
12、将pandas.Series()转换为指定shape的pandas.DataFrame
ser = pd.Series(np.random.randint(1, 10, 35))
df = pd.DataFrame(ser.values.reshape(7,5))
print(df)
13、取出pandas.Series()中满足条件数据的位置index
np.random.seed(666)
ser = pd.Series(np.random.randint(1, 10, 7))
print(ser)
np.argwhere(ser.values % 3 == 0)#数据为3的倍数的数据
14、取出pandas.Series()指定位置的数据
ser = pd.Series(list('abcdefghijklmnopqrstuvwxyz'))
pos = [0, 4, 8, 14, 20]#待取出位置
ser.take(pos)
15、pandas.Series()水平、垂直合并
ser1 = pd.Series(range(5))
ser2 = pd.Series(list('abcde'))
# 垂直合并
ser3 = ser1.append(ser2)
print(ser3)
# 水平合并
df = pd.concat([ser1, ser2], axis=1)
print(df)
16、输出pandas.Series()子集的index号
获取ser2在ser1中的index号
ser1 = pd.Series([10, 9, 6, 5, 3, 1, 12, 8, 13])
ser2 = pd.Series([1, 3, 10, 13])
# 方法1
[np.where(i == ser1)[0].tolist()[0] for i in ser2]
# 方法2
[pd.Index(ser1).get_loc(i) for i in ser2]
17、求真实和预测pd.Series之间的均方误差损失函数（MSE，mean squared error)
truth = pd.Series(range(10))
pred = pd.Series(range(10)) + np.random.random(10)
np.mean((truth-pred)**2)
18、pd.Series字符串型数据首字母大写转换
ser = pd.Series(['how', 'to', 'kick', 'ass?'])
# 方法一title()结合title()
ser.map(lambda x: x.title())
# 方法二upper()结合lambda
ser.map(lambda x: x[0].upper() + x[1:])
# 方法三title()结合列表推导式
pd.Series([i.title() for i in ser]) 
19、pd.Series字符串型数据字符长度计算
ser = pd.Series(['how', 'to', 'kick', 'ass?'])
ser.map(lambda x: len(x))
20、pd.Series中两两数之间差异
ser = pd.Series([1, 3, 6, 10, 15, 21, 27, 35])
print(ser.diff().tolist())
21、pd.Series中日期字符串转换为datetime格式
ser = pd.Series(['01 Jan 2010', '02-02-2011', '20120303', '2013/04/04', '2014-05-05', '2015-06-06T12:20'])
# 方法一
from dateutil.parser import parse
ser.map(lambda x: parse(x))
# 方法二
pd.to_datetime(ser)
22、获取pd.Series日期字符串中时间对象
ser = pd.Series(['01 Jan 2011', '02-02-2011', '20120303', '2013/04/04', '2014-05-05', '2015-06-06T12:20'])
from dateutil.parser import parse
ser_ts = ser.map(lambda x: parse(x))
# 处于该月中的那一天
print("Date: ", ser_ts.dt.day.tolist())
# week number
print("Week number: ", ser_ts.dt.weekofyear.tolist())
# 处于该年的第多少天
print("Day number of year: ", ser_ts.dt.dayofyear.tolist())
# 处于该年周几
print("Day of week: ", ser_ts.dt.day_name().tolist())#pythonic生物人修改，原文ser_ts.dt.weekday_name.tolist()有误
23、pd.Series日期字符串中修改为按指定日期输出
import pandas as pd
ser = pd.Series(['Jan 2010', 'Feb 2011', 'Mar 2012'])
# 方法1
from dateutil.parser import parse
# Parse the date
ser_ts = ser.map(lambda x: parse(x))
# Construct date string with date as 4
ser_datestr = ser_ts.dt.year.astype('str') + '-' + ser_ts.dt.month.astype('str') + '-' + '04'
# Format it.
[parse(i).strftime('%Y-%m-%d') for i in ser_datestr]
# 方法2
ser.map(lambda x: parse('04 ' + x))
24、输出pd.Series中至少包含两个元音字符的数据
ser = pd.Series(['Apple', 'Orange', 'Plan', 'Python', 'Money'])
from collections import Counter
mask = ser.map(lambda x: sum([Counter(x.lower()).get(i, 0) for i in list('aeiou')]) >= 2)
ser[mask]
25、输出pd.Series中有效的email地址
# 三种方法
emails = pd.Series(['buying books at amazom.com', '[email protected]', '[email protected]', '[email protected]'])
# Solution 1 (as series of strings)
import re
pattern ='[A-Za-z0-9._%+-]+@[A-Za-z0-9.-]+\\.[A-Za-z]{2,4}'
mask = emails.map(lambda x: bool(re.match(pattern, x)))
emails[mask]
# Solution 2 (as series of list)
emails.str.findall(pattern, flags=re.IGNORECASE)
# Solution 3 (as list)
[x[0] for x in [re.findall(pattern, email) for email in emails] if len(x) > 0]26、pd.Series1按pd.Series2分组并求均值
fruit = pd.Series(np.random.choice(['apple', 'banana', 'carrot'], 10))
weights = pd.Series(np.linspace(1, 10, 10))
weights.groupby(fruit).mean()
27、计算两个pd.Series之间的欧式距离
p = pd.Series([1, 2, 3, 4, 5, 6, 7, 8, 9, 10])
q = pd.Series([10, 9, 8, 7, 6, 5, 4, 3, 2, 1])
# 方法1
sum((p - q)**2)**.5
# 方法2
np.linalg.norm(p-q)
28、求pd.Series局部峰值index
ser = pd.Series([2, 10, 3, 4, 9, 10, 2, 7, 3])
dd = np.diff(np.sign(np.diff(ser)))
peak_locs = np.where(dd == -2)[0] + 1
peak_locs
29、pd.Series字符串数据中使用最低频字符填充空格
my_str = 'dbc deb abed gade'
ser = pd.Series(list('dbc deb abed gade'))
freq = ser.value_counts()
print(freq)
least_freq = freq.dropna().index[-1]
"".join(ser.replace(' ', least_freq))
30、创建时间序列数据，赋予随机值
ser = pd.Series(np.random.randint(1,10,10), pd.date_range('2000-01-01', periods=10, freq='W-SAT'))
ser
31、缺省的时间序列值 不同方式填充
ser = pd.Series([1,10,3, np.nan], index=pd.to_datetime(['2000-01-01', '2000-01-03', '2000-01-06', '2000-01-08']))
ser.resample('D').ffill() # 前填充
ser.resample('D').bfill() # f后填充
ser.resample('D').bfill().ffill() # 先后填充，如后无值则前填充
32、找出pd.Series中自相关性最大的数据
ser = pd.Series(np.arange(20) + np.random.normal(1, 10, 20))
autocorrelations = [ser.autocorr(i).round(2) for i in range(11)]
print(autocorrelations[1:])
print('Lag having highest correlation: ', np.argmax(np.abs(autocorrelations[1:]))+1) 
33、从一个csv 文件中每间隔50行取数据生成pandas.DataFrame
#三种方法
# Solution 1: Use chunks and for-loop
df = pd.read_csv('https://raw.githubusercontent.com/selva86/datasets/master/BostonHousing.csv', chunksize=50)
df2 = pd.DataFrame()
for chunk in df:
df2 = df2.append(chunk.iloc[0,:])
# Solution 2: Use chunks and list comprehension
df = pd.read_csv('https://raw.githubusercontent.com/selva86/datasets/master/BostonHousing.csv', chunksize=50)
df2 = pd.concat([chunk.iloc[0] for chunk in df], axis=1)
df2 = df2.transpose()
# Solution 3: Use csv reader
import csv
with open('BostonHousing.csv', 'r') as f:
reader = csv.reader(f)
out = []
for i, row in enumerate(reader):
if i%50 == 0:
out.append(row) 
34、从一个csv 文件取数据生成pandas.DataFrame（新增加一分类列）
##两种方式
# Solution 1: Using converter parameter
df = pd.read_csv('https://raw.githubusercontent.com/selva86/datasets/master/BostonHousing.csv',
converters={'medv': lambda x: 'High' if float(x) > 25 else 'Low'})
# Solution 2: Using csv reader
import csv
with open('BostonHousing.csv', 'r') as f:
reader = csv.reader(f)
out = []
for i, row in enumerate(reader):
if i > 0:
row[13] = 'High' if float(row[13]) > 25 else 'Low'
out.append(row)
df = pd.DataFrame(out[1:], columns=out[0])
print(df.head())
35、生成一个按规定步长平移的pandas.DataFrame
L = pd.Series(range(15))
def gen_strides(a, stride_len=5, window_len=5):
n_strides = ((a.size-window_len)//stride_len) + 1
return np.array([a[s:(s+window_len)] for s in np.arange(0, a.size, stride_len)[:n_strides]])
gen_strides(L, stride_len=2, window_len=4) 
36、从一个csv 文件读取指定列生成pandas.DataFrame
#usecols参数设置
df = pd.read_csv('https://raw.githubusercontent.com/selva86/datasets/master/BostonHousing.csv', usecols=['crim', 'medv'])
print(df.head()) 
37、输出DataFrame的行数、列数、数据类型、类型频数、Series转list
import pandas as pd
#https://raw.githubusercontent.com/selva86/datasets/master/Cars93_miss.csv
df = pd.read_csv('./Cars93_miss.csv')
# DataFrame的行数和列数
print(df.shape)
# DataFrame各列数据类型
print(df.dtypes.head())
# 每种数据类型的频数
print(df.dtypes.value_counts())
# DataFrame基础统计
df_stats = df.describe()
# pd.Series转list
df['Model'].tolist()[0]
38、输出满足某个规则的DataFrame数据行和列号
df = pd.read_csv('./Cars93_miss.csv')
#输出Price值最大时的行和列号
row, col = np.where(df.values == np.max(df.Price))#df.values输出DataFrame的值，为numpy.ndarray类型
print('%s\t%s'%(row,col))
39、修改DataFrame的列名称
#修改某一列的名称
df=df.rename(columns = {'Type':'CarType'})
# 或者
#df.columns.values[2] = "CarType"
# map函数修改某一匹配的名称：此处将名称中的.替换为下划线
df.columns = df.columns.map(lambda x: x.replace('.', '_'))
df.columns
40、DataFrame中是否有缺省值确认
df = pd.read_csv('./Cars93_miss.csv')
df.isnull().values.any()
41、DataFrame中缺省值统计
n_missings_each_col = df.apply(lambda x: x.isnull().sum())#每列缺省值计数
print(n_missings_each_col)
n_missings_each_col.argmax()#缺省值数最多列的列号
42、各自列均值填充DataFrame中各自列缺省值
df = pd.read_csv('./Cars93_miss.csv')
df_out = df[['Min.Price', 'Max.Price']] = df[['Min.Price', 'Max.Price']].apply(lambda x: x.fillna(x.mean()))
print(df_out.head())43、各自列均值、中值填充DataFrame中各自列缺省值（使用apply）
df = pd.read_csv('./Cars93_miss.csv')
d = {'Min.Price': np.nanmean, 'Max.Price': np.nanmedian}
df[['Min.Price', 'Max.Price']] = df[['Min.Price', 'Max.Price']].apply(lambda x, d: x.fillna(d[x.name](x)), args=(d, ))44、从DataFrame选择子DataFrame
df = pd.DataFrame(np.arange(20).reshape(-1, 5), columns=list('abcde'))
# 选择子DataFrame
type(df[['a']])
type(df.loc[:, ['a']])
type(df.iloc[:, [0]])
# 选择Series
type(df.a)
type(df['a'])
type(df.loc[:, 'a'])
type(df.iloc[:, 1])45、 改变DataFrame列顺序
import numpy as np
df = pd.DataFrame(np.arange(20).reshape(-1, 5), columns=list('abcde'))
# 方法1：传入一个list
df[list('cbade')]
# 方法2：自定义函数
def switch_columns(df, col1=None, col2=None):
colnames = df.columns.tolist()
i1, i2 = colnames.index(col1), colnames.index(col2)
colnames[i2], colnames[i1] = colnames[i1], colnames[i2]
return df[colnames]
df1 = switch_columns(df, 'a', 'c')
# 方法3：
df[sorted(df.columns)]
# 方法4：
df.sort_index(axis=1, ascending=False, inplace=True)46、大DataFrame修改默认显示的行和列数
df = pd.read_csv('./Cars93_miss.csv.txt')
pd.set_option('display.max_columns', 10)#输出10列
pd.set_option('display.max_rows', 10)#输出10行
df47、DataFrame数据小数位数设置
df = pd.DataFrame(np.random.random(4)**10, columns=['random'])
#方法1: Rounding
df.round(4)
#方法2: Use apply to change format
df.apply(lambda x: '%.4f' % x, axis=1)
# or
df.applymap(lambda x: '%.4f' % x)
#方法3: Use set_option
pd.set_option('display.float_format', lambda x: '%.4f' % x)
#方法4: Assign display.float_format
pd.options.display.float_format = '{:.4f}'.format
print(df)48、 DataFrame数据小数转百分比显示
df = pd.DataFrame(np.random.random(4), columns=['random'])
out = df.style.format({
'random': '{0:.2%}'.format,
})49、DataFrame数据每隔20行读取
df = pd.read_csv('./Cars93_miss.csv.txt')
print(df.iloc[::20, :][['Manufacturer', 'Model', 'Type']]) 
50、创建DataFrame主键
df = pd.read_csv('./Cars93_miss.csv', usecols=[0,1,2,3,5])
df[['Manufacturer', 'Model', 'Type']] = df[['Manufacturer', 'Model', 'Type']].fillna('missing')
df.index = df.Manufacturer + '_' + df.Model + '_' + df.Type
print(df.index.is_unique)51、获取DataFrame某一列中第n大的值索引
df = pd.DataFrame(np.random.randint(1, 30, 30).reshape(10,-1), columns=list('abc'))
n = 5#第5大
df['a'].argsort()[::-1][n]52、获取DataFrame某一列中第n大的值大于指定值得索引
ser = pd.Series(np.random.randint(1, 100, 15))
print('ser: ', ser.tolist(), 'mean: ', round(ser.mean()))
np.argwhere(ser > ser.mean())[1]
53、获取DataFrame中行和大于100的行
df = pd.DataFrame(np.random.randint(10, 40, 60).reshape(-1, 4))
# 求行和
rowsums = df.apply(np.sum, axis=1)#axis=1指定行
df.iloc[np.where(rowsums > 100)] 
54、 Series or DataFrame中使用分位数填充超限区域
#需求：使用0.05分位数和0.95分位数分别替换小于和大于该分位数的区域
ser = pd.Series(np.logspace(-2, 2, 30))
def cap_outliers(ser, low_perc, high_perc):
low, high = ser.quantile([low_perc, high_perc])#指定分位数
print(low_perc, '%ile: ', low, '|', high_perc, '%ile: ', high)
ser[ser < low] = low
ser[ser > high] = high
return(ser)
capped_ser = cap_outliers(ser, .05, .95)55、去除指定值将DataFrame转换为最大方阵
df = pd.DataFrame(np.random.randint(-20, 50, 100).reshape(10,-1))
print(df)
#去除负值
arr = df[df > 0].values.flatten()
arr_qualified = arr[~np.isnan(arr)]
#寻找最大可能的方阵维度
n = int(np.floor(arr_qualified.shape[0]**.5))
#方阵转换
top_indexes = np.argsort(arr_qualified)[::-1]
output = np.take(arr_qualified, sorted(top_indexes[:n**2])).reshape(n, -1)
print(output)
56、DataFrame两行交换
df = pd.DataFrame(np.arange(25).reshape(5, -1))
def swap_rows(df, i1, i2):
a, b = df.iloc[i1, :].copy(), df.iloc[i2, :].copy()
df.iloc[i1, :], df.iloc[i2, :] = b, a
return df
print(swap_rows(df, 1, 2))
57、DataFrame逆序输出
df = pd.DataFrame(np.arange(25).reshape(5, -1))
print(df)
#方法 1
df.iloc[::-1, :]
#方法 2
print(df.loc[df.index[::-1], :])
58、DataFrame转对角矩阵
df = pd.DataFrame(np.arange(25).reshape(5,-1), columns=list('abcde'))
print(df)
df_onehot = pd.concat([pd.get_dummies(df['a']), df[list('bcde')]], axis=1)
print(df_onehot)
59、DataFrame那一列含有最多行最大值
df = pd.DataFrame(np.random.randint(1,100, 40).reshape(10, -1))
print(df)
print('Column with highest row maxes: ', df.apply(np.argmax, axis=1).value_counts().index[0])
60、DataFrame创建新列：每行为行号（按欧几里得距离而来）
df = pd.DataFrame(np.random.randint(1,100, 40).reshape(10, -1), columns=list('pqrs'), index=list('abcdefghij'))
print(df)
import numpy as np
# init outputs
nearest_rows = []
nearest_distance = []
# iterate rows.
for i, row in df.iterrows():
curr = row
rest = df.drop(i)
e_dists = {} # init dict to store euclidean dists for current row.
# iterate rest of rows for current row
for j, contestant in rest.iterrows():
# compute euclidean dist and update e_dists
e_dists.update({j: round(np.linalg.norm(curr.values - contestant.values))})
# update nearest row to current row and the distance value
nearest_rows.append(max(e_dists, key=e_dists.get))
nearest_distance.append(max(e_dists.values()))
df['nearest_row'] = nearest_rows
df['dist'] = nearest_distance
df
61、求DataFrame各列之间最大相关系数
df = pd.DataFrame(np.random.randint(1,100, 80).reshape(8, -1), columns=list('pqrstuvwxy'), index=list('abcdefgh'))
print(df)
abs_corrmat = np.abs(df.corr())
max_corr = abs_corrmat.apply(lambda x: sorted(x)[-2])
print('Maximum Correlation possible for each column: ', np.round(max_corr.tolist(), 2))
62、DataFrame创建一列：包含每行中最小值与最大值比值
df = pd.DataFrame(np.random.randint(1,100, 80).reshape(8, -1))
print(df)
#方法 1
min_by_max = df.apply(lambda x: np.min(x)/np.max(x), axis=1)
#方法 2
min_by_max = np.min(df, axis=1)/np.max(df, axis=1)
min_by_max
63、DataFrame创建一列：包含每行中第二大的值
df = pd.DataFrame(np.random.randint(1,100, 80).reshape(8, -1))
out = df.apply(lambda x: x.sort_values().unique()[-2], axis=1)
df['penultimate'] = out
df
64、DataFrame每列按特定方式归一化
df = pd.DataFrame(np.random.randint(1,100, 80).reshape(8, -1))
# 均值、标准差归一化
out1 = df.apply(lambda x: ((x - x.mean())/x.std()).round(2))
print('均值、标准差归一化：\n',out1)
# 最大最小值归一化
out2 = df.apply(lambda x: ((x.max() - x)/(x.max() - x.min())).round(2))
print('最大最小值归一化：\n', out2) 
65、计算DataFrame每行与后一行的相关系数
df = pd.DataFrame(np.random.randint(1,100, 80).reshape(8, -1))
[df.iloc[i].corr(df.iloc[i+1]).round(2) for i in range(df.shape[0])[:-1]]
66、DataFrame对角线元素替换为0
df = pd.DataFrame(np.random.randint(1,100, 100).reshape(10, -1))
print(df)
for i in range(df.shape[0]):
df.iat[i, i] = 0
df.iat[df.shape[0]-i-1, i] = 0
df
67、DataFrame按某列分组、提取某个分组
df = pd.DataFrame({'col1': ['apple', 'banana', 'orange'] * 3,
'col2': np.random.rand(9),
'col3': np.random.randint(0, 15, 9)})
print(df)
df_grouped = df.groupby(['col1'])#按col1列分组
#获取指定分组，此处取apple组
# 方法 1
df_grouped.get_group('apple')
# 方法 2
for i, dff in df_grouped:
if i == 'apple':
print(dff)
68、DataFrame按另外列分组、提取当前列中指定值（看下方例子，需求不好描述）
df = pd.DataFrame({'fruit': ['apple', 'banana', 'orange'] * 3,
'taste': np.random.rand(9),
'price': np.random.randint(0, 15, 9)})
print(df)
df_grpd = df['taste'].groupby(df.fruit)#taste值按fruit类别排序
df_grpd.get_group('banana').sort_values().iloc[-2]#获取fruit中的banana组，取第二大值
69、DataFrame分组（看下方例子，需求不好描述）
df = pd.DataFrame({'fruit': ['apple', 'banana', 'orange'] * 3,
'rating': np.random.rand(9),
'price': np.random.randint(0, 15, 9)})
print(df)
out = df.groupby('fruit', as_index=False)['price'].mean()#按fruit分组、求每类fruit的均价
print(out)
70、两个DataFrame使用类似SQL 中INNER JOIN拼接
# Join dataframes df1 and df2 by ‘fruit-pazham’ and ‘weight-kilo’.
df1 = pd.DataFrame({
'fruit': ['apple', 'banana', 'orange'] * 3,
'weight': ['high', 'medium', 'low'] * 3,
'price': np.random.randint(0, 15, 9)
})
df2 = pd.DataFrame({
'pazham': ['apple', 'orange', 'pine'] * 2,
'kilo': ['high', 'low'] * 3,
'price': np.random.randint(0, 15, 6)
})
print('df1:\n')
print(df1)
print('df2:\n')
print(df2)
# Solution
pd.merge(df1,
df2,
how='inner',
left_on=['fruit', 'weight'],
right_on=['pazham', 'kilo'],
suffixes=['_left', '_right'])
71、移除DataFrame1中在DataFrame2出现的行
df1 = pd.DataFrame({
'fruit': ['apple', 'orange', 'banana'] * 3,
'weight': ['high', 'medium', 'low'] * 3,
'price': np.arange(9)
})
df2 = pd.DataFrame({
'fruit': ['apple', 'orange', 'pine'] * 2,
'weight': ['high', 'medium'] * 3,
'price': np.arange(6)
})
print('df1:\n')
print(df1)
print('df2:\n')
print(df2)
#
print(df1[~df1.isin(df2).all(1)])
72、取出DataFrame中两列值相等的行号
df = pd.DataFrame({
'fruit1': np.random.choice(['apple', 'orange', 'banana'], 10),
'fruit2': np.random.choice(['apple', 'orange', 'banana'], 10)
})
print(df)
np.where(df.fruit1 == df.fruit2)
73、DataFrame中新建两列：滞后列和提前列（看下方例子，需求BT）
df = pd.DataFrame(np.random.randint(1, 100, 20).reshape(-1, 4),
columns=list('abcd'))
print(df)
df['a_lag1'] = df['a'].shift(1)#a列每行值向后移一位
df['b_lead1'] = df['b'].shift(-1)#b列每行值向前移一位
print(df) 
74、DataFrame中所有值出现频次统计
df = pd.DataFrame(np.random.randint(1, 10, 20).reshape(-1, 4),
columns=list('abcd'))
pd.value_counts(df.values.ravel())
75、拆分DataFrame中某列文本为两列
df = pd.DataFrame([
"STD, City State", "33, Kolkata West Bengal",
"44, Chennai Tamil Nadu", "40, Hyderabad Telengana",
"80, Bangalore Karnataka"
],
columns=['row'])
print(df)
df_out = df.row.str.split(',|\t', expand=True)#拆分
new_header = df_out.iloc[0]#第一列设置为标题
df_out = df_out[1:]
df_out.columns = new_header
print(df_out)