Multi-Label_Text_classification_for_Household_grocery_Items

.py Files:

preprocessing.py

• It converts the labels in training data in one hot encoding
• Splits the train_data into train.csv and validation.csv and saves the csvs in the Data folder

dataloader.py –

• I have used torchtext to tokenize the titles, creating the vocabulary and creating the trainloader, validationloader and testloader

model.py –

• It contain the Bi-lstm network architecture

train_and_save.py –

• This script is used for training the network and then saving the checkpoints of the network

test.py –

• This script is used to test the network on the unseen data.
• In the Results folder -> test_results.csv is generated using test.py ( Data used is test.csv which is saved inside Data folder )

Model Used:

• I attempted to solve the multi-label classification problem using 2 layer Bi-LSTM . ( For the full network configuration see model.py)
• For the Loss Function I have used BCEWithLogitsLoss
• Optimizer used : Adam

Final Loss after 12 Epoches of training:

Training Loss: 0.0049 Validation Loss: 0.0028

Model_Checkpoints folder contains the weight files. Final result csv is saved in the Results folder -> test_results.csv. It contains the titles and probability scores for each of the labels.

Packages Used:

PyTorch, Sklearn, Torchtext, Pandas, Numpy

Built With

• python
• pytorch