时间戳

首页 > 解决方案 > Pandas：将带有 if/else 条件的 for 循环转换为 apply 方法（lambda 函数）

python - Pandas：将带有 if/else 条件的 for 循环转换为 apply 方法（lambda 函数）

问题描述

我有以下带有 for 循环的功能：

def add_CQI_iterrows(df):
    previous_row = df['Date'].astype(str)[0]
    CQI_index = 0
    series = []

    for index, row in df.iterrows():
        if row['Date'] == previous_row:       
            previous_row = row['Date']
            print(CQI_index)
        else:
            CQI_index += 1
            previous_row = row['Date']      
        series.append(CQI_index)
    df['CQI'] = series
    
    return df

我想找到一种方法将此 for 循环转换为 apply 方法。像这样的东西（不起作用）：

def add_CQI_apply(df):
    previous_row = df['Date'].astype(str)[0]
    CQI_index = 1
    series = []
    
    df['CQI'] = df.apply(lambda row: previous_row = row['Date'] if row['Date'] == previous_row else CQI_index += 1 and previous_row = row['Date'], axis=1)
    
    return df

我想做这个转换，因为我想看看 apply 方法有多快，以及是否可以在 Pandas 系列上对 apply 方法进行矢量化。

这是我的数据（data.json）：

[
 {
   "Date": "9/20/2020 8:50",
   "UE": 1
 },
 {
   "Date": "9/20/2020 8:50",
   "UE": 2
 },
 {
   "Date": "9/20/2020 8:50",
   "UE": 3
 },
 {
   "Date": "9/20/2020 8:57",
   "UE": 1
 },
 {
   "Date": "9/20/2020 8:57",
   "UE": 8
 },
 {
   "Date": "9/20/2020 8:57",
   "UE": 2
 },
 {
   "Date": "9/20/2020 9:12",
   "UE": 1
 },
 {
   "Date": "9/20/2020 9:12",
   "UE": 5
 },
 {
   "Date": "9/20/2020 9:12",
   "UE": 3
 },
 {
   "Date": "9/20/2020 9:20",
   "UE": 1
 },
 {
   "Date": "9/20/2020 9:20",
   "UE": 4
 },
 {
   "Date": "9/20/2020 9:20",
   "UE": 3
 }
]

最后是上传这些数据的函数：

def upload_data(file):
    df = pd.read_json(file)
    df['Date'] = pd.to_datetime(df['Date'], format="%Y-%d-%m %H:%M:%S") 
    df['CQI'] = np.nan
    return df

标签： pythonpandasfor-loopvectorizationapply

解决方案

df['CQI'] = (df['Date'] != df['Date'].shift()).cumsum()

In [120]: (df['Date'] != df['Date'].shift()).cumsum()
Out[120]:
0     1
1     1
2     1
3     2
4     2
5     2
6     3
7     3
8     3
9     4
10    4
11    4
Name: Date, dtype: int64

推荐阅读