将数据读取到numpy数组

Question

I have a file below

label,feature
0,70 80 90 50 33 58 ...
2,53 56 84 56 25 12 ...
1,32 56 84 89 65 87 ...
...
2,56 48 57 56 99 22 ...
4,25 65 84 54 54 15 ...

I want the data could be

Ytrain = [0,2,1,...2,4]  (int, ndarray)
Xtrain = [[70 80 90 50 33 58...],
          [53 56 80 56 25 12...],
          ...
          [25 65 84 54 54 15...]] (int, ndarray)

here is my code

data = pd.read_csv('train.csv')
Ytrain = np.array(data.iloc[:, 0]).astype(int)
train = np.array(data.iloc[:, 1:]).astype(str)

Xtrain = []
for i in range(len(train)):
    tmp = [int(x) for x in train[i][0].split()]
    Xtrain.append(tmp)
Xtrain = np.array(Xtrain)

do you have a better way to do that ?

Answer 1

Add multiple separator to read_csv with header=None and skiprows=1 for not read csv header:

data = pd.read_csv('train.csv', sep="[,\s+]", header=None, skiprows=1, engine='python')
print (data)
   0   1   2   3   4   5   6
0  0  70  80  90  50  33  58
1  2  53  56  84  56  25  12
2  1  32  56  84  89  65  87
3  2  56  48  57  56  99  22
4  4  25  65  84  54  54  15

Last select by iloc :

Ytrain = data.iloc[:,0].values
Xtrain = data.iloc[:,1:].values

---

Or use split with expand=True for DataFrame :

data = pd.read_csv('train.csv')
Ytrain = data.iloc[:,0].values.astype(int)
Xtrain = data.iloc[:,1].str.split(expand=True).values.astype(int)

print (Ytrain)
[0 2 1 2 4]

print (Xtrain)
[[70 80 90 50 33 58]
 [53 56 84 56 25 12]
 [32 56 84 89 65 87]
 [56 48 57 56 99 22]
 [25 65 84 54 54 15]]

Answer 2

You can use numpy for this. Since you have multiple delimiters, a little more work is required.

import numpy as np

s = open('train.csv', 'r').read().replace(',', ' ')
arr = np.genfromtxt(s)

Ytrain = arr[:, 1]
Xtrain = arr[:, 1:]