Digikala Comments Analysis with RNN - LSTM

Predicting whether a comment left in Digikala without any idea about the recommendation, is recommending a product or not.

• The project overview
  • Data Preprocessing
    • Cleaning
    • Converting
    • Checking null
    • Analyzing
    • Normalizing
  • Modeling
    • Tokenizing
    • Vectorizing
    • Creating model
    • Training model
    • Evaluating model

Example results

model.predict(vectorize_layer(['خیلی بد بود']))
array([[0.23936771]], dtype=float32)

model.predict(vectorize_layer(['افتضاح بود']))
array([[0.04372201]], dtype=float32)

model.predict(vectorize_layer(['عالی بود']))
array([[0.99864143]], dtype=float32)

model.predict(vectorize_layer(['من از این محصول رضایت دارم']))
array([[0.99523926]], dtype=float32)

model.predict(vectorize_layer(['با این قیمت ارزش خرید ندارد']))
array([[0.11104872]], dtype=float32)

model.predict(vectorize_layer(['با این قیمت ارزش خرید دارد']))
array([[0.9712803]], dtype=float32)

License

This project is licensed under the MIT License - see the LICENSE.md file for details

by : Shahab Rahnama