CODEWITHSUNDEEP
pythonpandasrolling-computation
freddy888
freddy888

Reputation: 1010

Pandas: Rolling 2nd largest value

I need to get the rolling 2nd largest value of a df.

To get the largest value I do

max = df.sort_index(ascending=True).rolling(10).max()

When I try this, python throws an error

max = df.sort_index(ascending=True).rolling(10).nlargest(2)

AttributeError: 'Rolling' object has no attribute 'nlargest'

Is this a bug? What else can I use that is performant?

Upvotes: 3

Views: 1867

Answers (2)

jezrael
jezrael

Reputation: 863226

Use np.sort in descending order and select second value:

np.random.seed(2019)

df = pd.DataFrame({
    'B': np.random.randint(20, size=15)
})
print (df)
     B
0    8
1   18
2    5
3   15
4   12
5   10
6   16
7   16
8    7
9    5
10  19
11  12
12  16
13  18
14   5

a = df.rolling(10).apply(lambda x: -np.sort(-x)[1]) 
#alternative
#a = df.rolling(10).apply(lambda x: np.sort(x)[-2]) 
print (a)
       B
0    NaN
1    NaN
2    NaN
3    NaN
4    NaN
5    NaN
6    NaN
7    NaN
8    NaN
9   16.0
10  18.0
11  16.0
12  16.0
13  18.0
14  18.0

Upvotes: 3

Itamar Mushkin
Itamar Mushkin

Reputation: 2905

I'd do something like this:

df.rolling(10).apply(lambda x: pd.Series(x).nlargest(2).iloc[-1])

Upvotes: 6

Related Questions