Amazon Sales Predictor

This is a small learning project where I train a PyTorch model to estimate a fair Amazon product price in rupees.

The model uses:

• rating
• discount percentage
• category id

Then it predicts a price.

Data

The notebook downloads this Kaggle dataset with kagglehub: karkavelrajaj/amazon-sales-dataset

Notebook Content

1. Load and clean the data (discounted_price, discount_percentage, rating).
2. Convert category text to a numeric category_id.
3. Apply log + min-max normalization on price.
4. Build tensors and split into train/test (80/20).
5. Train a neural network (2 hidden layers, ReLU, MSELoss, Adam).
6. Check test loss and plot predicted vs actual.
7. Enter new values and get a predicted fair price.

How to run

1. Install packages:

pip install kagglehub pandas numpy matplotlib torch

2. Open sales_predictor.ipynb.
3. Run cells from top to bottom.
4. In the last cell, enter:

• rating (1 to 5)
• discount percentage (0 to 100)
• category id (shown range in notebook)