Pandas 计算同比环比指标的3种方法 - 《Pandas 数据分析从入门到实战》

发表评论
12161 views

 ›
 ›
Pandas 计算同比环比指标的3种方法 - 《Pandas 数据分析从入门到实战》



同比和环比：环比和同比用于描述统计数据的变化情况

  • 环比：表示本次统计段与相连的上次统计段之间的比较。
    • 比如 2010 年中国第一季度 GDP 为 G2010Q1 亿元，第二季度 GDP 为 G2010Q2 亿元，则第二季度 GDP 环比增长（G2010Q2-G2010Q1)/G2010Q1；
  • 同比：即同期相比，表示某个特定统计段今年与去年之间的比较。
    • 比如 2009 年中国第一季度 GDP 为 G2009Q1 亿元，则 2010 年第一季度的 GDP 同比增长为（G2010Q1-G2009Q1)/G2009Q1。


演示： 读取连续3年的天气数据

  • 方法1：pandas.Series.pct_change
  • 方法2：pandas.Series.shift
  • 方法3：pandas.Series.diff

pct_change、shift、diff，都实现了跨越多行的数据计算

读取连续 3 年的天气数据


import pandas as pd
%matplotlib inline

fpath = "./datas/beijing_tianqi/beijing_tianqi_2017-2019.csv"
df = pd.read_csv(fpath, index_col="ymd", parse_dates=True)

df.head(3)
           bWendu yWendu tianqi fengxiang fengli  aqi aqiInfo  aqiLevel
ymd
2017-01-01  5℃     -3℃    霾~晴    南风        1-2级   450  严重污染    6
2017-01-02  7℃     -6℃    晴~霾    南风        1-2级   246  重度污染    5
2017-01-03  5℃     -5℃    霾      南风        1-2级   320  严重污染    6


# 替换掉温度的后缀 ℃
df["bWendu"] = df["bWendu"].str.replace("℃", "").astype('int32')

df.head(3)
            bWendu yWendu tianqi fengxiang fengli  aqi aqiInfo  aqiLevel
ymd
2017-01-01  5       -3℃    霾~晴    南风        1-2级   450  严重污染    6
2017-01-02  7       -6℃    晴~霾    南风        1-2级   246  重度污染    5
2017-01-03  5       -5℃    霾      南风        1-2级   320  严重污染    6


# 新的df，为每个月的平均最高温
df = df[["bWendu"]].resample("M").mean()
# 将索引按照日期升序排列
df.sort_index(ascending=True, inplace=True)

df.head()
               bWendu
ymd
2017-01-31  3.322581
2017-02-28  7.642857
2017-03-31  14.129032
2017-04-30  23.700000
2017-05-31  29.774194
df.index
DatetimeIndex(['2017-01-31', '2017-02-28', '2017-03-31', '2017-04-30',
               '2017-05-31', '2017-06-30', '2017-07-31', '2017-08-31',
               '2017-09-30', '2017-10-31', '2017-11-30', '2017-12-31',
               '2018-01-31', '2018-02-28', '2018-03-31', '2018-04-30',
               '2018-05-31', '2018-06-30', '2018-07-31', '2018-08-31',
               '2018-09-30', '2018-10-31', '2018-11-30', '2018-12-31',
               '2019-01-31', '2019-02-28', '2019-03-31', '2019-04-30',
               '2019-05-31', '2019-06-30', '2019-07-31', '2019-08-31',
               '2019-09-30', '2019-10-31', '2019-11-30', '2019-12-31'],
              dtype='datetime64[ns]', name='ymd', freq='M')


df.plot()


方法1：pandas.Series.pct_change

pct_change 方法直接算好了"(新-旧)/旧"的百分比

官方文档地址：https://pandas.pydata.org/pandas-docs/stable/reference/api/pandas.Series.pct_change.html

df["bWendu_way1_huanbi"] = df["bWendu"].pct_change(periods=1)
df["bWendu_way1_tongbi"] = df["bWendu"].pct_change(periods=12)

df.head(15)
               bWendu  bWendu_way1_huanbi  bWendu_way1_tongbi
ymd
2017-01-31  3.322581  NaN                 NaN
2017-02-28  7.642857   1.300277           NaN
2017-03-31  14.129032  0.848658           NaN
2017-04-30  23.700000  0.677397           NaN
2017-05-31  29.774194  0.256295           NaN
2017-06-30  30.966667  0.040051           NaN
2017-07-31  31.612903  0.020869           NaN
2017-08-31  30.129032 -0.046939           NaN
2017-09-30  27.866667 -0.075089           NaN
2017-10-31  17.225806 -0.381849           NaN
2017-11-30  9.566667  -0.444632           NaN
2017-12-31  4.483871  -0.531303           NaN
2018-01-31  1.322581  -0.705036           -0.601942
2018-02-28  4.892857   2.699477           -0.359813
2018-03-31  14.129032  1.887685            0.000000

方法2：pandas.Series.shift

shift用于移动数据，但是保持索引不变

官方文档地址：https://pandas.pydata.org/pandas-docs/stable/reference/api/pandas.Series.shift.html

# 见识一下shift做了什么事情
# 使用pd.concat合并Series列表变成一个大的df
pd.concat(
    [df["bWendu"], 
     df["bWendu"].shift(periods=1), 
     df["bWendu"].shift(periods=12)],
    axis=1
).head(15)
               bWendu     bWendu     bWendu
ymd
2017-01-31  3.322581  NaN        NaN
2017-02-28  7.642857   3.322581  NaN
2017-03-31  14.129032  7.642857  NaN
2017-04-30  23.700000  14.129032 NaN
2017-05-31  29.774194  23.700000 NaN
2017-06-30  30.966667  29.774194 NaN
2017-07-31  31.612903  30.966667 NaN
2017-08-31  30.129032  31.612903 NaN
2017-09-30  27.866667  30.129032 NaN
2017-10-31  17.225806  27.866667 NaN
2017-11-30  9.566667   17.225806 NaN
2017-12-31  4.483871   9.566667  NaN
2018-01-31  1.322581   4.483871   3.322581
2018-02-28  4.892857   1.322581   7.642857
2018-03-31  14.129032  4.892857   14.129032
# 环比
series_shift1 = df["bWendu"].shift(periods=1)
df["bWendu_way2_huanbi"] = (df["bWendu"]-series_shift1)/series_shift1

# 同比
series_shift2 = df["bWendu"].shift(periods=12)
df["bWendu_way2_tongbi"] = (df["bWendu"]-series_shift2)/series_shift2
df.head(15)
               bWendu  bWendu_way1_huanbi  ...  bWendu_way2_huanbi  bWendu_way2_tongbi
ymd                                        ...
2017-01-31  3.322581  NaN                  ... NaN                 NaN
2017-02-28  7.642857   1.300277            ...  1.300277           NaN
2017-03-31  14.129032  0.848658            ...  0.848658           NaN
2017-04-30  23.700000  0.677397            ...  0.677397           NaN
2017-05-31  29.774194  0.256295            ...  0.256295           NaN
2017-06-30  30.966667  0.040051            ...  0.040051           NaN
2017-07-31  31.612903  0.020869            ...  0.020869           NaN
2017-08-31  30.129032 -0.046939            ... -0.046939           NaN
2017-09-30  27.866667 -0.075089            ... -0.075089           NaN
2017-10-31  17.225806 -0.381849            ... -0.381849           NaN
2017-11-30  9.566667  -0.444632            ... -0.444632           NaN
2017-12-31  4.483871  -0.531303            ... -0.531303           NaN
2018-01-31  1.322581  -0.705036            ... -0.705036           -0.601942
2018-02-28  4.892857   2.699477            ...  2.699477           -0.359813
2018-03-31  14.129032  1.887685            ...  1.887685            0.000000

[15 rows x 5 columns]


方法3. pandas.Series.diff

pandas.Series.diff 用于新值减去旧值

官方文档：https://pandas.pydata.org/pandas-docs/stable/reference/api/pandas.Series.diff.html

pd.concat(
    [df["bWendu"], 
     df["bWendu"].diff(periods=1), 
     df["bWendu"].diff(periods=12)],
    axis=1
).head(15)
               bWendu     bWendu  bWendu
ymd
2017-01-31  3.322581  NaN        NaN
2017-02-28  7.642857   4.320276  NaN
2017-03-31  14.129032  6.486175  NaN
2017-04-30  23.700000  9.570968  NaN
2017-05-31  29.774194  6.074194  NaN
2017-06-30  30.966667  1.192473  NaN
2017-07-31  31.612903  0.646237  NaN
2017-08-31  30.129032 -1.483871  NaN
2017-09-30  27.866667 -2.262366  NaN
2017-10-31  17.225806 -10.640860 NaN
2017-11-30  9.566667  -7.659140  NaN
2017-12-31  4.483871  -5.082796  NaN
2018-01-31  1.322581  -3.161290  -2.00
2018-02-28  4.892857   3.570276  -2.75
2018-03-31  14.129032  9.236175   0.00


# 环比
series_diff1 = df["bWendu"].diff(periods=1)
df["bWendu_way3_huanbi"] = series_diff1/(df["bWendu"]-series_diff1)

# 同比
series_diff2 = df["bWendu"].diff(periods=12)
df["bWendu_way3_tongbi"] = series_diff2/(df["bWendu"]-series_diff2)
df.head(15)
               bWendu  bWendu_way1_huanbi  ...  bWendu_way2_huanbi  bWendu_way2_tongbi
ymd                                        ...
2017-01-31  3.322581  NaN                  ... NaN                 NaN
2017-02-28  7.642857   1.300277            ...  1.300277           NaN
2017-03-31  14.129032  0.848658            ...  0.848658           NaN
2017-04-30  23.700000  0.677397            ...  0.677397           NaN
2017-05-31  29.774194  0.256295            ...  0.256295           NaN
2017-06-30  30.966667  0.040051            ...  0.040051           NaN
2017-07-31  31.612903  0.020869            ...  0.020869           NaN
2017-08-31  30.129032 -0.046939            ... -0.046939           NaN
2017-09-30  27.866667 -0.075089            ... -0.075089           NaN
2017-10-31  17.225806 -0.381849            ... -0.381849           NaN
2017-11-30  9.566667  -0.444632            ... -0.444632           NaN
2017-12-31  4.483871  -0.531303            ... -0.531303           NaN
2018-01-31  1.322581  -0.705036            ... -0.705036           -0.601942
2018-02-28  4.892857   2.699477            ...  2.699477           -0.359813
2018-03-31  14.129032  1.887685            ...  1.887685            0.000000

[15 rows x 5 columns]