Question

我有一个按比赛情况分类的NBA球员数据集，我想知道是否有一种方法可以获取通常用于连续值的 mode 统计等效项，但要获取发生频率最高的字符串值？

    t1_start1           t1_start2          t1_start3       t1_start4    t1_start5   team1
0   Shaquille O'Neal    Kobe Bryant         Horace Grant    Ron Harper  Rick Fox    LAL
1   Shaquille O'Neal    Kobe Bryant         Horace Grant    Ron Harper  Rick Fox    LAL
2   Kobe Bryant         Shaquille O'Neal    Horace Grant    Ron Harper  Brian Shaw  LAL
3   Kobe Bryant         Shaquille O'Neal    Horace Grant    Brian Shaw  Ron Harper  LAL
4   Kobe Bryant         Shaquille O'Neal    Horace Grant    Ron Harper  Brian Shaw  LAL
5   LeBron James        Brandon Ingram      Kyle Kuzma      JaVale McGeeLonzo  Ball  LAL

不管玩家入门者的订单(t1_start1 | t1_start2 | t1_start3 | ... )如何，我如何在过去3行中按“ team1”列分组来排名前5位的最常用玩家？

Answer 1

您可以将np.unique()与return_counts=True和np.argsort()结合使用：

players, starts = np.unique(df[['t1_start1','t1_start2','t1_start3','t1_start4','t1_start5']].values, return_counts=True)

players[np.argsort(-starts)][:5]

返回：

['Horace Grant' 'Kobe Bryant' 'Ron Harper' "Shaquille O'Neal" 'Brian Shaw']

Answer 2

flat_list = df.loc[[0:3]].values.flatten() # first 3 rows flattened to a 1d list
print(scipy.stats.mode(flat_list).mode) # the most common element in that list

如果您想要多个值，可以使用collections.Counter

most_common_5 = collection.Counter(flat_list).most_common(5)

获取最频繁出现的字符串值？

2 个答案: