# How to handle imbalance dataset on text classification

**URL:** <https://discuss.pytorch.org/t/how-to-handle-imbalance-dataset-on-text-classification/97724>\
**Category:** nlp\
**Created:** [September 29, 2020, 12:46pm UTC](https://discuss.pytorch.org/t/how-to-handle-imbalance-dataset-on-text-classification/97724 "2020-09-29T12:46:09Z")\
**Posts on this page:** 2\
**Page:** 1

<div class="post-metadata">

**Author:** ![smi222](https://discuss.pytorch.org/letter_avatar_proxy/v4/letter/s/67e7ee/32.png) [@smi222](https://discuss.pytorch.org/u/smi222)\
**Post date:** [September 29, 2020, 12:46pm UTC](https://discuss.pytorch.org/t/how-to-handle-imbalance-dataset-on-text-classification/97724/1 "2020-09-29T12:46:09Z")

</div>

Hello Everyone,

I am working on binary text classification problem. How do i apply smote or WeightedRandomSample for the imbalance in my dataset.My code currenlty looks like this:

```auto
class GDataset(Dataset):

  def __init__ (self, passage, targets, tokenizer, max_len):
    self.passage = passage
    self.targets = targets
    self.tokenizer = tokenizer
    self.max_len = max_len
  
  def __len__ (self):
    return len(self.passage)
  
  def __getitem__ (self, item):
    passage = str(self.passage[item])
    target = self.targets[item]
    
    if (target == 1) and self.transform: # minority class
            x = self.transform(x)

    encoding = self.tokenizer.encode_plus(
      passage,
      add_special_tokens=True,
      max_length=self.max_len,
      return_token_type_ids=False,
      pad_to_max_length=True,
      return_attention_mask=True,
      return_tensors='pt',
    )
    return
      'passage_text': passage,
      'input_ids': encoding['input_ids'].flatten(),
      'attention_mask': encoding['attention_mask'].flatten(),
      'targets': torch.tensor(target, dtype=torch.long

```

How can i apply other balancing techniques

---

<div class="post-metadata">

**Author:** ![ptrblck](https://discuss.pytorch.org/user_avatar/discuss.pytorch.org/ptrblck/32/1823_2.png) [@ptrblck](https://discuss.pytorch.org/u/ptrblck)\
**Post date:** [September 30, 2020, 9:14am UTC](https://discuss.pytorch.org/t/how-to-handle-imbalance-dataset-on-text-classification/97724/2 "2020-09-30T09:14:24Z")

</div>

If each sample contains a single target value, you should be able to directly use the `WeightedRandomSampler` as described in [this example](https://discuss.pytorch.org/t/how-to-handle-imbalanced-classes/11264/2).