I am trying to build an algorithm for finding number of clusters. I need to assign random points from a data set as initial means.
I first tried the following code :
mu=random.sample(df,10)
it gave index out of range error.
I converted the same into a numpy array and then did
mu=random.sample(np.array(df).tolist(),10)
instead of giving 10 values as mean it is giving me 10 arrays of values.
How can I get a 10 values to initialise as mean for 10 clusters from the dataframe?
I think you need DataFrame.sample:
mu = df.sample(10)
Sample:
np.random.seed(100)
df = pd.DataFrame(np.random.randint(10, size=(20,3)), columns=list('abc'))
print (df)
a b c
0 8 8 3
1 7 7 0
2 4 2 5
3 2 2 2
4 1 0 8
5 4 0 9
6 6 2 4
7 1 5 3
8 4 4 3
9 7 1 1
10 7 7 0
11 2 9 9
12 3 2 5
13 8 1 0
14 7 6 2
15 0 8 2
16 5 1 8
17 1 5 4
18 2 8 3
19 5 0 9
mu = df.sample(10)
print (mu)
a b c
11 2 9 9
1 7 7 0
8 4 4 3
5 4 0 9
2 4 2 5
19 5 0 9
13 8 1 0
14 7 6 2
0 8 8 3
9 7 1 1
Use numpy.random.choice
df.iloc[np.random.choice(np.arange(len(df)), 10, False)]
Or numpy.random.permutation
df.loc[np.random.permutation(df.index)[:10]]
a b c
11 2 9 9
1 7 7 0
16 5 1 8
15 0 8 2
17 1 5 4
19 5 0 9
10 7 7 0
8 4 4 3
6 6 2 4
14 7 6 2
