Python 使用Pandas查找具有重复列值的行并将其聚合的最佳方法

我创建了一个函数，用于查找具有一组相同列值的行。不过这很难看：嵌套的apply函数。是否有更好的方法来测试n个数字列是否具有相同的值，如果是，则对其应用函数或将其添加到字典

#function to add valid polls to dictionary
def j(x,i,cand_name):
        if (x['_pollname'] in valid_polls):
            if (x['_poll'] in valid_elections):
                if(np.isfinite(x[cand_name])):
                    toappenddf = {}
                    toappenddf['candidate'] = str(cand_name)
                    toappenddf['date'] = x['_date']
                    toappenddf['poll'] = x['_pollname']
                    toappenddf['polled'] = x[cand_name]
                    appendc.append(toappenddf)

invalid_cols = ['_poll','_pollname','_date']

# run the function
for val in enumerate(list(df3.columns.values)):
    if (val[1] not in invalid_cols):
        df3.apply(lambda row: j(row,i,val[1]), axis=1)

#create dataframe from results
df0 = pd.DataFrame(appendc)
pollaggregator = []

#create nested for loops
def k(x):
    def h(w,i,xx):
        # see if there's a match along three column values between any two rows
        if ((w.candidate == xx.candidate) & (w.date == xx.date) & (xx.poll != w.poll)):
            appendpoll = {}
            appendpoll = w.tolist()
            pollaggregator.append(appendpoll)
    df0.apply(lambda row: h(row,i,x), axis=1)

    if (len(pollaggregator) > 0):
        appendpoll2 = {}
        appendpoll2 = x.tolist()

df0.apply(k, axis=1)

输入数据是数据帧中的轮询数据，如下所示：

,candidate,date,poll,polled
0,Biden,2014-12-21 00:00:00,CNN,8.0
1,Biden,2015-01-27 00:00:00,FOX News,17.0
2,Biden,2015-02-15 00:00:00,CNN,14.0
3,Biden,2015-03-02 00:00:00,Quinnipiac,10.0
4,Biden,2015-03-13 00:00:00,CNN/ORC,15.0
5,Biden,2015-03-31 00:00:00,FOX News,12.0
6,Biden,2015-04-24 00:00:00,FOX News,9.0
7,Biden,2015-04-20 00:00:00,CNN/Opinion Research,11.0
8,Biden,2015-04-23 00:00:00,Quinnipiac,10.0
9,Biden,2015-04-24 00:00:00,FOX News,9.0

---

出于我的目的，我希望聚合具有相同日期值但不同轮询名称值的行。

示例输入和输出通常会有所帮助。因此，您可以按日期进行唯一分组吗？@PadraicCunningham是的，但只能根据某些标准-例如。，如果轮询名称相同，但轮询组织不同…@Incodeveritas是否可以上载至少包含几百行的示例文件？例如，dropbox与您的csv文件共享链接。所以轮询和日期相等，但轮询必须唯一？基本上是df.groupbydate、poll和其中的唯一元素？