How to Split/Partition a Dataset into Training and Test Datasets For, E.G., Cross Validation

How to split/partition a dataset into training and test datasets for, e.g., cross validation?

If you want to split the data set once in two parts, you can use numpy.random.shuffle, or numpy.random.permutation if you need to keep track of the indices (remember to fix the random seed to make everything reproducible):

import numpy
# x is your dataset
x = numpy.random.rand(100, 5)
numpy.random.shuffle(x)
training, test = x[:80,:], x[80:,:]

or

import numpy
# x is your dataset
x = numpy.random.rand(100, 5)
indices = numpy.random.permutation(x.shape[0])
training_idx, test_idx = indices[:80], indices[80:]
training, test = x[training_idx,:], x[test_idx,:]

There are many ways other ways to repeatedly partition the same data set for cross validation. Many of those are available in the sklearn library (k-fold, leave-n-out, ...). sklearn also includes more advanced "stratified sampling" methods that create a partition of the data that is balanced with respect to some features, for example to make sure that there is the same proportion of positive and negative examples in the training and test set.

How do I split a custom dataset into training and test datasets?

Using Pytorch's SubsetRandomSampler:

import torch
import numpy as np
from torchvision import datasets
from torchvision import transforms
from torch.utils.data.sampler import SubsetRandomSampler

class CustomDatasetFromCSV(Dataset):
    def __init__(self, csv_path, transform=None):
        self.data = pd.read_csv(csv_path)
        self.labels = pd.get_dummies(self.data['emotion']).as_matrix()
        self.height = 48
        self.width = 48
        self.transform = transform

    def __getitem__(self, index):
        # This method should return only 1 sample and label 
        # (according to "index"), not the whole dataset
        # So probably something like this for you:
        pixel_sequence = self.data['pixels'][index]
        face = [int(pixel) for pixel in pixel_sequence.split(' ')]
        face = np.asarray(face).reshape(self.width, self.height)
        face = cv2.resize(face.astype('uint8'), (self.width, self.height))
        label = self.labels[index]

        return face, label

    def __len__(self):
        return len(self.labels)

dataset = CustomDatasetFromCSV(my_path)
batch_size = 16
validation_split = .2
shuffle_dataset = True
random_seed= 42

# Creating data indices for training and validation splits:
dataset_size = len(dataset)
indices = list(range(dataset_size))
split = int(np.floor(validation_split * dataset_size))
if shuffle_dataset :
    np.random.seed(random_seed)
    np.random.shuffle(indices)
train_indices, val_indices = indices[split:], indices[:split]

# Creating PT data samplers and loaders:
train_sampler = SubsetRandomSampler(train_indices)
valid_sampler = SubsetRandomSampler(val_indices)

train_loader = torch.utils.data.DataLoader(dataset, batch_size=batch_size, 
                                           sampler=train_sampler)
validation_loader = torch.utils.data.DataLoader(dataset, batch_size=batch_size,
                                                sampler=valid_sampler)

# Usage Example:
num_epochs = 10
for epoch in range(num_epochs):
    # Train:   
    for batch_index, (faces, labels) in enumerate(train_loader):
        # ...

How to split a dataset with multiple variables into train and test while both having the same composition using python?

If you want to get specific rows from a pandas DataFrame that meet certain criteria, you can filter. In your case, something like:

reader_lung = reader[reader["Image_Title"] == "Lung"]

"Image_Title" you need to change to the name of the column you're looking for your keyword (e.g., Lung) in. This needs to be an exact match.

For something that doesn't require an exact match, you could also do the following:

reader_lung = reader[reader["Image_Title"].str.contains("Lung")]

How to split data into 3 sets (train, validation and test)?

Numpy solution. We will shuffle the whole dataset first (df.sample(frac=1, random_state=42)) and then split our data set into the following parts:

• 60% - train set,
• 20% - validation set,
• 20% - test set

---

In [305]: train, validate, test = \
              np.split(df.sample(frac=1, random_state=42), 
                       [int(.6*len(df)), int(.8*len(df))])

In [306]: train
Out[306]:
          A         B         C         D         E
0  0.046919  0.792216  0.206294  0.440346  0.038960
2  0.301010  0.625697  0.604724  0.936968  0.870064
1  0.642237  0.690403  0.813658  0.525379  0.396053
9  0.488484  0.389640  0.599637  0.122919  0.106505
8  0.842717  0.793315  0.554084  0.100361  0.367465
7  0.185214  0.603661  0.217677  0.281780  0.938540

In [307]: validate
Out[307]:
          A         B         C         D         E
5  0.806176  0.008896  0.362878  0.058903  0.026328
6  0.145777  0.485765  0.589272  0.806329  0.703479

In [308]: test
Out[308]:
          A         B         C         D         E
4  0.521640  0.332210  0.370177  0.859169  0.401087
3  0.333348  0.964011  0.083498  0.670386  0.169619

[int(.6*len(df)), int(.8*len(df))] - is an indices_or_sections array for numpy.split().

Here is a small demo for np.split() usage - let's split 20-elements array into the following parts: 80%, 10%, 10%:

In [45]: a = np.arange(1, 21)

In [46]: a
Out[46]: array([ 1,  2,  3,  4,  5,  6,  7,  8,  9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20])

In [47]: np.split(a, [int(.8 * len(a)), int(.9 * len(a))])
Out[47]:
[array([ 1,  2,  3,  4,  5,  6,  7,  8,  9, 10, 11, 12, 13, 14, 15, 16]),
 array([17, 18]),
 array([19, 20])]

Splitting dataset into two non-redundant numpy arrays?

Following this answer you can do:

train_idx = np.random.randint(my_data.shape[0], size=split_size)    
mask = np.ones_like(my_data, dtype=bool)
mask[train_idx] = False
train, test = my_data[~mask], my_data[mask]

Although, a more natural way would be to slice a permutation of your data, as Poojan suggested.

permuted = np.random.permutation(my_data)
train, test = permuted[:split_size], permuted[split_size:]

Is there a rule-of-thumb for how to divide a dataset into training and validation sets?

There are two competing concerns: with less training data, your parameter estimates have greater variance. With less testing data, your performance statistic will have greater variance. Broadly speaking you should be concerned with dividing data such that neither variance is too high, which is more to do with the absolute number of instances in each category rather than the percentage.

If you have a total of 100 instances, you're probably stuck with cross validation as no single split is going to give you satisfactory variance in your estimates. If you have 100,000 instances, it doesn't really matter whether you choose an 80:20 split or a 90:10 split (indeed you may choose to use less training data if your method is particularly computationally intensive).

Assuming you have enough data to do proper held-out test data (rather than cross-validation), the following is an instructive way to get a handle on variances:

1. Split your data into training and testing (80/20 is indeed a good starting point)
2. Split the training data into training and validation (again, 80/20 is a fair split).
3. Subsample random selections of your training data, train the classifier with this, and record the performance on the validation set
4. Try a series of runs with different amounts of training data: randomly sample 20% of it, say, 10 times and observe performance on the validation data, then do the same with 40%, 60%, 80%. You should see both greater performance with more data, but also lower variance across the different random samples
5. To get a handle on variance due to the size of test data, perform the same procedure in reverse. Train on all of your training data, then randomly sample a percentage of your validation data a number of times, and observe performance. You should now find that the mean performance on small samples of your validation data is roughly the same as the performance on all the validation data, but the variance is much higher with smaller numbers of test samples

---

Related Topics