In the field of data science, pandas is a very useful tool. In the part of big data (usually related to deep learning) in the field of data science, this blog starts from the important functions of pandas, to data transformation and data analysis. Pandas provides major data processing functions such as data transformation, data cleaning, data visualization, and data extraction. The python version of the blog used in this article is:
#!/usr/local/bin/python3
# Copyright 2022 shichaog
import platform
print(platform.python_version())
输入：
3.8.7
pandas-basics
To use pandas in python, you need to import this module. Use import pandas as pd, dataframe is the most commonly used data format in pandas. Stock data, supermarket department store transaction data, etc. can be represented by dataframe. The dataframe is created as follows:
#!/usr/local/bin/python3
# Copyright 2022 shichaog
import pandas as pd
Speeds={
'Animal': ['Falcon', 'Falcon',
'Parrot', 'Parrot'],
'MaxSpeed': [380., 370., 24., 23.0]}
df = pd.DataFrame(Speeds)
print(df)
其输出是一张表，如下：
Animal MaxSpeed
0 Falcon 380.0
1 Falcon 370.0
2 Parrot 24.0
3 Parrot 23.0
In the above code, Speeds is a dictionary, Animal and MaxSpeed are key values, followed by list is the corresponding value value, 0, 1, 2, 3 are index values, which are used for dataframe index, and can be used for retrieval, query, modification, operations such as deletion.
If you want to get a column of values, you only need to add square brackets after the dataframe name, and write the column name in the square brackets, as shown below:
#!/usr/local/bin/python3
# Copyright 2022 shichaog
import pandas as pd
df = pd.DataFrame({
'Animal': ['Falcon', 'Falcon',
'Parrot', 'Parrot'],
'MaxSpeed': [380., 370., 24., 23.0]})
#下一行的等价写法是：print(df.MaxSpeed)
print(df['MaxSpeed'])
输出为：
0 380.0
1 370.0
2 24.0
3 23.0
Pandas can also add columns, filter and other operations:
#接上df
#增加一列
df['Size'] = [16., 60., 25., 40.]
print(df)
输出为：
Animal MaxSpeed Size
0 Falcon 380.0 16.0
1 Falcon 370.0 60.0
2 Parrot 24.0 25.0
3 Parrot 23.0 40.0
#筛选过滤
print(df[df['Size'] >= 40.0])
输出为：
Animal MaxSpeed Size
1 Falcon 370.0 60.0
3 Parrot 23.0 40.0
In addition to building the dataframe by yourself, you can also read in some existing file formats, such as csv, xsl, etc. iris.csv is the address of the iris dataset, a kaggle open source dataset commonly used in deep learning, and you can use the pd.read_csv method . The basic usage of pandas dataframe is as follows:
#!/usr/local/bin/python3
# Copyright 2022 shichaog
import pandas as pd
iris = pd.read_csv('iris.csv')
#数据样本信息
print(iris.shape)
输出：
(150, 5)
#样本实例
print(iris.head(5))
print(iris.tail(5))
输出：
sepallength sepalwidth petallength petalwidth class
0 5.1 3.5 1.4 0.2 Iris-setosa
1 4.9 3.0 1.4 0.2 Iris-setosa
2 4.7 3.2 1.3 0.2 Iris-setosa
3 4.6 3.1 1.5 0.2 Iris-setosa
4 5.0 3.6 1.4 0.2 Iris-setosa
sepallength sepalwidth petallength petalwidth class
145 6.7 3.0 5.2 2.3 Iris-virginica
146 6.3 2.5 5.0 1.9 Iris-virginica
147 6.5 3.0 5.2 2.0 Iris-virginica
148 6.2 3.4 5.4 2.3 Iris-virginica
149 5.9 3.0 5.1 1.8 Iris-virginica
#数据类型
print(iris.dtypes)
输出：
sepallength float64
sepalwidth float64
petallength float64
petalwidth float64
class object
dtype: object
#数据取子集，选择3，,4，,5三个行，行索引从0开始
print(iris.loc[3:5])
输出：
sepallength sepalwidth petallength petalwidth class
3 4.6 3.1 1.5 0.2 Iris-setosa
4 5.0 3.6 1.4 0.2 Iris-setosa
5 5.4 3.9 1.7 0.4 Iris-setosa
#相比上面，增加了列，信息，列可以是列名，也可以是列所以，如果是是列索引值，则需要使用iloc
#下面两个作用等同，都是选取3行0列所在位置的值
print(iris.loc[3, 'sepallength'])
print(iris.iloc[3,0])
输出：
4.6
4.6
#数据导出为csv
iris.to_csv('iris-out.csv', index=False)
In addition, pandas also has some processing methods for abnormal data, which will be put in the following sections.
pandas data calculation
data type conversion
Here we take NASA's planetary data as an example,
#!/usr/local/bin/python3
# Copyright 2022 shichaog
import pandas as pd
planets = pd.read_csv('planets.csv')
print(planets.head(3))
输出
method number orbital_period mass distance year
0 Radial Velocity 1 269.300 7.10 77.40 2006
1 Radial Velocity 1 874.774 2.21 56.95 2008
2 Radial Velocity 1 763.000 2.60 19.84 2011
print(planets.dtypes)
输出
method object
number int64
orbital_period float64
mass float64
distance float64
year int64
dtype: object
#获取均值，所有结果都是浮点数
print(planets.mean())
输出
number 1.785507
orbital_period 2002.917596
mass 2.638161
distance 264.069282
year 2009.070531
dtype: float64
#整数除以浮点数，结果是浮点数
print(planets['number'][0]/planets['mass'][0])
输出
0.14084507042253522
#使用astype强制类型转换，将整数类型转换为浮点数类型
print(planets['number'][0].astype(float))
输出
1.0
#强制类型转换，将浮点数转换为整数
print(planets['mass'][0].astype(int))
输出
7
planets['year'][0].astype(str)
输出
‘2006’
planets['year_dt'] = pd.to_datetime(planets['year'], format='%Y')
print(planets['year_dt'])
输出
0 2006-01-01
1 2008-01-01
2 2011-01-01
3 2007-01-01
4 2009-01-01
...
1030 2006-01-01
1031 2007-01-01
1032 2007-01-01
1033 2008-01-01
1034 2008-01-01
Name: year_dt, Length: 1035, dtype: datetime64[ns]
string type
In pandas, .strit is a string accessor, which provides a large number of string manipulation methods.
#!/usr/local/bin/python3
# Copyright 2022 shichaog
import pandas as pd
names = pd.Series([' github; Shichaog','csdn; Shichaog'])
#字符串替换，将;替换成/
names = names.str.replace(';','/')
print(names)
#字符串长度获取
print(names.str.len())
#删除字符串前的前导空格
names = names.str.strip()
print(names)
print(names.str.len())
#字符串大小写转换，.lower是大写转小写
names = names.str.upper()
print(names)
#按;分割series为list
names = names.str.split('; ')
print(names)
#::-1是list索引方法
names = pd.Series([i[::-1] for i in names])
print(names)
#jion方法连接单词
names = [' '.join(i) for i in names]
print(names)
Date data processing
#!/usr/local/bin/python3
# Copyright 2022 shichaog
import pandas as pd
#构建日期period_range生成日期series，第一个参数是起始日期，第二个参数生成频率
daterange = pd.period_range('1/1/2020', freq='30d', periods=4)
date_df = pd.DataFrame(data=daterange,columns=['sample date'])
print(date_df)
输出：
sample date
0 2020-01-01
1 2020-01-31
2 2020-03-01
3 2020-03-31
#日期差异，使用diff，period是比较周期，在股票趋势派交易中常会比较均线和股价的上下穿关系，也常用到diff
date_df['date difference'] = date_df['sample date'].diff(periods=1)
print(date_df)
输出：
sample date date difference
0 2020-01-01 NaT
1 2020-01-31 <30 * Days>
2 2020-03-01 <30 * Days>
3 2020-03-31 <30 * Days>
#查询该月第一天
date_df['first of month'] = date_df['sample date'].values.astype('datetime64[M]')
print(date_df)
输出：
sample date date difference first of month
0 2020-01-01 NaT 2020-01-01
1 2020-01-31 <30 * Days> 2020-01-01
2 2020-03-01 <30 * Days> 2020-03-01
3 2020-03-31 <30 * Days> 2020-03-01
#数据类型
print(date_df.dtypes)
输出：
sample date period[30D]
date difference object
first of month datetime64[ns]
dtype: object
date_df['sample date'] = date_df['sample date'].dt.to_timestamp()
print(date_df.dtypes)
输出：
sample date datetime64[ns]
date difference object
first of month datetime64[ns]
dtype: object
#数据相减
date_df['sample date'] - date_df['first of month']
date_df['sample date'] - date_df['date difference']
date_df['sample date'] - pd.Timedelta('30 d')
#使用dt获取更多属性
date_df['sample date'].dt.day_name()
Error data handling
Uncleaned data will have errors, missing, etc. According to statistics, it usually takes 80%-90% of the time for data cleaning in the entire project. pandas and python provide some data cleaning methods, which can greatly save data cleaning. time
numerical error
For numerical data, such as trading volume, there are three types of errors that may occur: value loss, value error, and data duplication.
pandas.isnull can be used to judge the missing data. The following array is a missing value when it is initialized with nan, and this method can be used to judge.
#!/usr/local/bin/python3
# Copyright 2022 shichaog
>>>array = np.array([[1, np.nan, 3], [4, 5, np.nan]])
>>>array
array([[ 1., nan, 3.],
[ 4., 5., nan]])
>>>pd.isna(array)
array([[False, True, False],
[False, False, True]])
>>>index = pd.DatetimeIndex(["2017-07-05", "2017-07-06", None,
"2017-07-08"])
>>>index
DatetimeIndex(['2017-07-05', '2017-07-06', 'NaT', '2017-07-08'],
dtype='datetime64[ns]', freq=None)
>>>pd.isna(index)
array([False, False, True, False])
>>>df = pd.DataFrame([['ant', 'bee', 'cat'], ['dog', None, 'fly']])
>>>df
0 1 2
0 ant bee cat
1 dog None fly
>>>pd.isna(df)
0 1 2
0 False False False
1 False True False
In addition, methods such as dropna() can be used to remove non-numeric rows/columns, or the fillna() method can be set to a certain value.
outlier
This type of value is beyond the normal range, such as cpu usage, car speed, human height, weight and other data all have a reasonable range, which is more in line with the normal distribution in statistics.
#!/usr/local/bin/python3
# Copyright 2022 shichaog
import pandas as pd
#这里的1000和1.0是两个异常值
df = pd.DataFrame({
'Animal': ['Falcon', 'Falcon','Falcon', 'Falcon', 'Falcon',
'Parrot', 'Parrot', 'Parrot', 'Parrot', 'Parrot'],
'MaxSpeed': [380., 370., 330., 1000, 320, 24., 26., 23.0, 1.0, 25.0]})
#查看数据统计信息
print(df.groupby('Animal').describe())
输出：
MaxSpeed
count mean std min 25% 50% 75% max
Animal
Falcon 5.0 480.0 291.804729 320.0 330.0 370.0 380.0 1000.0
Parrot 5.0 19.8 10.568822 1.0 23.0 24.0 25.0 26.0
#统计样本数
print(df['Animal'].value_counts())
输出
Falcon 5
Parrot 5
#见下图所绘制
pd.pivot(df, columns='Animal').plot(subplots=True)
#通过以下句子筛选错误值
print(df.query('Animal=="Falcon" & ( MaxSpeed > 400.)'))
From the figure above, we can see that there are obviously two values far beyond the average value, which often need to be pre-processed before training the model.
apply & map
The method provided by functions such as apply and map can be used to modify the pandas dataframe value. The advantage of this type of method is that it is no longer necessary to use statements such as for/loop in other languages.
#!/usr/local/bin/python3
# Copyright 2022 shichaog
import pandas as pd
#read in apple stock info
apple_df = pd.read_csv('../AAPL Historical Data.csv')
该df的输出如下，其是一个股票的daily交易信息，包括日期、收盘价、开盘价以及成交量和换手率等。
Date Price Open High Low Vol. Change %
0 Jul 19, 2022 151.00 147.98 151.20 146.92 82.15M 2.67%
1 Jul 18, 2022 147.07 150.84 151.54 146.74 77.54M -2.06%
2 Jul 15, 2022 150.17 149.78 150.86 148.20 76.26M 1.15%
3 Jul 14, 2022 148.47 144.08 148.95 143.25 78.14M 2.05%
4 Jul 13, 2022 145.49 142.99 146.45 142.12 71.19M -0.25%
5 Jul 12, 2022 145.86 145.76 148.45 145.05 77.59M 0.68%
6 Jul 11, 2022 144.87 145.67 146.64 143.78 63.31M -1.48%
7 Jul 08, 2022 147.04 145.26 147.55 145.00 64.30M 0.47%
8 Jul 07, 2022 146.35 143.29 146.55 143.28 65.73M 2.40%
9 Jul 06, 2022 142.92 141.35 144.12 141.08 73.55M 0.96%
10 Jul 05, 2022 141.56 137.77 141.61 136.93 70.95M 1.89%
11 Jul 01, 2022 138.93 136.04 139.04 135.66 71.05M 1.62%
12 Jun 30, 2022 136.72 137.25 138.37 133.77 98.63M -1.80%
13 Jun 29, 2022 139.23 137.46 140.67 136.67 65.98M 1.30%
14 Jun 28, 2022 137.44 142.13 143.42 137.32 66.75M -2.98%
15 Jun 27, 2022 141.66 142.70 143.49 140.96 70.21M 0.00%
16 Jun 24, 2022 141.66 139.90 141.91 139.77 88.44M 2.45%
17 Jun 23, 2022 138.27 136.82 138.59 135.63 72.11M 2.16%
18 Jun 22, 2022 135.35 134.79 137.76 133.91 73.12M -0.38%
19 Jun 21, 2022 135.87 133.42 137.06 133.32 80.68M 3.28%
#drop uncessary
apple_df=apple_df.drop(columns=['Open', 'High', 'Low', 'Vol.', 'Change %'])
#reverse and calculate Moving Average
apple_df = apple_df.iloc[::-1]
apple_df['SMA3'] = apple_df['Price'].rolling(3).mean()
Date Price SMA3
19 Jun 21, 2022 135.87 NaN
18 Jun 22, 2022 135.35 NaN
17 Jun 23, 2022 138.27 136.496667
16 Jun 24, 2022 141.66 138.426667
15 Jun 27, 2022 141.66 140.530000
14 Jun 28, 2022 137.44 140.253333
13 Jun 29, 2022 139.23 139.443333
12 Jun 30, 2022 136.72 137.796667
11 Jul 01, 2022 138.93 138.293333
10 Jul 05, 2022 141.56 139.070000
9 Jul 06, 2022 142.92 141.136667
8 Jul 07, 2022 146.35 143.610000
7 Jul 08, 2022 147.04 145.436667
6 Jul 11, 2022 144.87 146.086667
5 Jul 12, 2022 145.86 145.923333
4 Jul 13, 2022 145.49 145.406667
3 Jul 14, 2022 148.47 146.606667
2 Jul 15, 2022 150.17 148.043333
1 Jul 18, 2022 147.07 148.570000
0 Jul 19, 2022 151.00 149.413333
#drop na moving average
apple_df = apple_df.dropna()
#apply method to alter values along an axis
apple_df['Cross_direction'] = apple_df.apply(lambda x: 'upper' if x['Price']>x['SMA3'] else 'lower',axis=1)
Date Price SMA3 Cross_direction
17 Jun 23, 2022 138.27 136.496667 upper
16 Jun 24, 2022 141.66 138.426667 upper
15 Jun 27, 2022 141.66 140.530000 upper
14 Jun 28, 2022 137.44 140.253333 lower
13 Jun 29, 2022 139.23 139.443333 lower
12 Jun 30, 2022 136.72 137.796667 lower
11 Jul 01, 2022 138.93 138.293333 upper
10 Jul 05, 2022 141.56 139.070000 upper
9 Jul 06, 2022 142.92 141.136667 upper
8 Jul 07, 2022 146.35 143.610000 upper
7 Jul 08, 2022 147.04 145.436667 upper
6 Jul 11, 2022 144.87 146.086667 lower
5 Jul 12, 2022 145.86 145.923333 lower
4 Jul 13, 2022 145.49 145.406667 upper
3 Jul 14, 2022 148.47 146.606667 upper
2 Jul 15, 2022 150.17 148.043333 upper
1 Jul 18, 2022 147.07 148.570000 lower
0 Jul 19, 2022 151.00 149.413333 upper
#map method to substitute each value in a series
cross_map = {
"upper":"Red","lower":"Blue"}
apple_df['Cross Color'] = apple_df['Cross_direction'].map(cross_map)
Date Price SMA3 Cross_direction Cross Color
17 Jun 23, 2022 138.27 136.496667 upper Red
16 Jun 24, 2022 141.66 138.426667 upper Red
15 Jun 27, 2022 141.66 140.530000 upper Red
14 Jun 28, 2022 137.44 140.253333 lower Blue
13 Jun 29, 2022 139.23 139.443333 lower Blue
12 Jun 30, 2022 136.72 137.796667 lower Blue
11 Jul 01, 2022 138.93 138.293333 upper Red
10 Jul 05, 2022 141.56 139.070000 upper Red
9 Jul 06, 2022 142.92 141.136667 upper Red
8 Jul 07, 2022 146.35 143.610000 upper Red
7 Jul 08, 2022 147.04 145.436667 upper Red
6 Jul 11, 2022 144.87 146.086667 lower Blue
5 Jul 12, 2022 145.86 145.923333 lower Blue
4 Jul 13, 2022 145.49 145.406667 upper Red
3 Jul 14, 2022 148.47 146.606667 upper Red
2 Jul 15, 2022 150.17 148.043333 upper Red
1 Jul 18, 2022 147.07 148.570000 lower Blue
0 Jul 19, 2022 151.00 149.413333 upper Red
applymap_df=apple_df.applymap(lambda x: len(str(x)))
Date Price SMA3 Cross_direction Cross Color
17 12 6 18 5 3
16 12 6 18 5 3
15 12 6 6 5 3
14 12 6 18 5 4
13 12 6 18 5 4
12 12 6 18 5 4
11 12 6 18 5 3
10 12 6 6 5 3
9 12 6 18 5 3
8 12 6 18 5 3
7 12 6 18 5 3
6 12 6 18 5 4
5 12 6 18 5 4
4 12 6 18 5 3
3 12 6 18 5 3
2 12 6 18 5 3
1 12 6 18 5 4
0 12 5 18 5 3
print(apple_df)
Dataframe transformation
grouping and aggregation
It can be achieved by groupby and agg methods.
#!/usr/local/bin/python3
# Copyright 2022 shichaog
import pandas as pd
iris = pd.read_csv('iris.csv')
print(iris.head(5))
sepallength sepalwidth petallength petalwidth class
0 5.1 3.5 1.4 0.2 Iris-setosa
1 4.9 3.0 1.4 0.2 Iris-setosa
2 4.7 3.2 1.3 0.2 Iris-setosa
3 4.6 3.1 1.5 0.2 Iris-setosa
4 5.0 3.6 1.4 0.2 Iris-setosa
#根据花类别分组，然后使用max方法聚合
print(iris.groupby(['class']).max())
sepallength sepalwidth petallength petalwidth
class
Iris-setosa 5.8 4.4 1.9 0.6
Iris-versicolor 7.0 3.4 5.1 1.8
Iris-virginica 7.9 3.8 6.9 2.5
#通过.agg传递多个聚合参数，参数是字典形式的
df = iris.groupby(['class']).agg({
'petallength':['mean','min','max'],'petalwidth':'count'})
print(df)
petallength petalwidth
mean min max count
class
Iris-setosa 1.464 1.0 1.9 50
Iris-versicolor 4.260 3.0 5.1 50
Iris-virginica 5.552 4.5 6.9 50
#聚合之后，可以用下面方法修改列名
df.columns = ['_'.join(col).strip() for col in df.columns.values]
df.reset_index()
print(df)
petallength_mean ... petalwidth_count
class ...
Iris-setosa 1.464 ... 50
Iris-versicolor 4.260 ... 50
Iris-virginica 5.552 ... 50
groupings = iris.groupby(['class'])
groupings.get_group('Iris-setosa').head()
print(groupings.max())
sepallength sepalwidth petallength petalwidth
class
Iris-setosa 5.8 4.4 1.9 0.6
Iris-versicolor 7.0 3.4 5.1 1.8
Iris-virginica 7.9 3.8 6.9 2.5
#可以使用lamda方法，这里同groupings.max()作用是一样的
groupings.apply(lambda x: x.max())
sepallength sepalwidth ... petalwidth class
class ...
Iris-setosa 5.8 4.4 ... 0.6 Iris-setosa
Iris-versicolor 7.0 3.4 ... 1.8 Iris-versicolor
Iris-virginica 7.9 3.8 ... 2.5 Iris-virginica
#这是使用lamda方法过滤出最大值小于5的信息。
groupings.filter(lambda x: x['petalwidth'].max() <5)
sepallength sepalwidth petallength petalwidth class
0 5.1 3.5 1.4 0.2 Iris-setosa
1 4.9 3.0 1.4 0.2 Iris-setosa
2 4.7 3.2 1.3 0.2 Iris-setosa
3 4.6 3.1 1.5 0.2 Iris-setosa
4 5.0 3.6 1.4 0.2 Iris-setosa
.. ... ... ... ... ...
reshape
Sometimes it is necessary to transform the original pandas dataframe before using it. The four commonly used methods in pandas to realize this function are stack(), unstack(), pivot() and melt().
pivot
Pivot is used to generate a new dataframe, regenerate a DataFrame object with a given index (index), column (column) and given values (Values). This function does not support data aggregation, multiple values will result in a multi-index on the column.
pivot(index=None,columns=None,values=None) -> DataFrame
index: Specify a column as the index to generate the DataFrame object, if it is empty, it will default to the original index.
columns: Specify the value of a column as the column name, and the value must be passed.
values: Specify a column as the value of the generated DataFrame object. Can be empty.
#!/usr/local/bin/python3
# Copyright 2022 shichaog
import pandas as pd
data = {
'水果':['苹果','梨','草莓','苹果','梨','草莓'],
'商店':["C1","C1","C1", "C2","C2","C2"],
'价格':[10,9,8,8,6,8]}
df = pd.DataFrame(data)
print(df)
输出：
水果 商店 价格
0 苹果 C1 10
1 梨 C1 9
2 草莓 C1 8
3 苹果 C2 8
4 梨 C2 6
5 草莓 C2 8
df_pivot = df.pivot(index='水果',columns='商店',values='价格')
print(df_pivot)
输出：
商店 C1 C2
水果
梨 9 6
苹果 10 8
草莓 8 8
stack和unstack
stack provides the opposite function of pivot, turning columns into rows.
df2 = df.set_index(['水果','商店'])
print(df2)
stacked_df = pd.DataFrame(df2.stack())
print(stacked_df)
unstack_df = stacked_df.unstack('商店')
Then the above df2 and stacked_df are as follows:
df2
stacked_df
uses unstack
melt
melt_df = df.melt(id_vars=['水果','商店'], var_name='value type')
print(melt_df)
pivot_table
pivot_table: Reshape the data through the specified index and column, which can be aggregated.
pivot_table_df = df.pivot_table(index='水果',columns='商店',values='价格')
print(pivot_table_df)
splicing and merging
merge
The merge method is used to merge two dataframes and series into one dataframe. The keyword how indicates the method of merging, and on indicates which field to merge.
df1 = pd.DataFrame({
'Char': ['A', 'B', 'C', 'D'],
'number': [1, 2, 3, 4]})
df2 = pd.DataFrame({
'Char': ['C', 'D', 'E', 'F'],
'number': [3, 4, 5, 6]})
merge_df = df1.merge(df2,how='left',on='number')
inner_df = df1.merge(df2,how='inner',left_on='number',right_on='number')
m2_df = df1.merge(df2,how='right',on='number',suffixes=('','_right'))
Since df2 does not have two fields A and B, its value is nan after merging.
concat and join
#可以用drop_duplicates去掉重复的内容
df3 = pd.concat([df1,df2]).drop_duplicates().reset_index(drop=True)
#水平方向拼接
df4 = pd.concat([df1,df2],axis=1)
#在dataframe尾部增加
new_row = pd.Series(['Z',26],index=df3.columns)
df3.append(new_row,ignore_index=True)
#根据索引拼接
join_df = pd.DataFrame({
'Char': ['F','G', 'H', 'I'],
'number': [6, 7, 8, 9]})
df2.join(join_df, rsuffix='_right')
drawing
Pandas provides a variety of drawing tools, which can draw linear graphs, histograms, and relationship matrix graphs, etc.
df.plot();
df.plot.area(stacked=True);
df.hist();
from pandas.plotting import scatter_matrix
scatter_matrix(df,figsize=(4, 6),);
In addition, in a subdivided industry or field, there are also some dedicated drawing packages, such as seanborn for drawing heat maps, mplfinance for drawing financial data maps such as stocks, etc.
Statistics
#均值
df.mean()
#中位数
df.median()
.mode()
.std()
#数据概览快速方法
.describe()
#数据本身相关方法
.corr()