1. Removed puntuation, extra spaces and digits
2. Lower case all text
3. Chose the first Node of product_category_tree to be Main Category
4. Removed Categories with < 100 rows

1. Catorical plots using Seaborn and matplotlib
2. Stiplots of price, discounts vs Category
3. Reviews Anlysis
4. Word Cloud

1. Initially used a DistillBERT but switched to XLNet because it performs better at classification and has no number of token
2. Used a pretrained XLNet adding a Classifier layer on top of it
3. Balanced class imbalance using class weights
4. finally used description with name, brand, and product specification

Final Validation F1 Score: 74

Confusion Matrix:

• Using a data augmentation method to fix class imbalance
  • using translation
  • using hypernyms and hyponymns and synonyms