amirsoahil101/Flipkart_Product_Model

🛍️ Flipkart Product Review Sentiment & Rating Predictor (NLP)

An End-to-End Natural Language Processing (NLP) web application that predicts product ratings (Rate: 1 to 5) based on customer reviews and feedback summaries using TF-IDF tokenization and supervised Machine Learning classifiers.

---

Next-Generation Vocal Acoustic Analytics for Emergency Triage Scoring

---

🖼️ Application Preview

---

🔗 Live Application

Access the interactive live web application:
👉 Click Here to Launch Live Application

---

💡 Model Selection & Engineering Challenges

During the development and experimental phase, multiple model architectures were evaluated:

❌ Artificial Neural Network (ANN) Limitations

• High Loss & Accuracy Fluctuations: Training an ANN architecture led to severe epoch-to-epoch instability in both accuracy and training loss.
• Validation Loss Stagnation: The test/validation loss failed to converge effectively, indicating that deep neural layers were struggling to generalize on the sparse TF-IDF text matrix without overfitting.
• Decision: The ANN approach was discarded due to unpredictable performance and poor generalization on unseen test data.

🏆 Machine Learning Approach (Selected Solution)

• Stable Generalization: Transitioned to traditional Machine Learning classification algorithms which provided consistent convergence.
• High Test Performance: The final trained ML model achieved an 83% Test Accuracy, successfully resolving the variance/loss issues observed in the ANN model while ensuring fast, reliable inference for real-time web deployment.

---

⚙️ Data Preprocessing & NLP Pipeline

The input dataset (flipkart_product.csv) combines Review and Summary into a single feature string Text. Custom text normalization is applied prior to tokenization:

1. Punctuation Removal: Cleans standard punctuation marks using str.maketrans.
2. Lowercasing: Standardizes characters to lowercase.
3. Digit Removal: Strips numeric digits from input text.
4. Emoji & Non-ASCII Removal: Filters non-ASCII characters to retain clean English tokens.
5. Vectorization: Transforms normalized text into numeric matrices using the serialized tokenizer.pkl.

---

📁 Repository Structure

├── flipkart_product.csv   # Primary Product Reviews Dataset
├── .gitignore             # Git Ignore File
├── .gitattributes         # Git Attributes File
├── requirements.txt       # Python Dependencies
├── notebook.ipynb         # EDA, Text Normalization & Model Experiments
├── app.py                 # Interactive Streamlit Web Application
├── model.pkl              # Serialized Trained Machine Learning Classifier (83% Test Acc)
├── tokenizer.pkl          # Fitted TF-IDF / Vectorizer Object
├── columns.pkl            # Feature Column Schema Definitions
├── UI.png                 # Application UI Preview
└── README.md              # Documentation

---

git clone https://github.com/amirsohail100/Flipkart-Product-Review-Sentiment-Rating-Predictor-NLP-.git

cd Flipkart-Product-Review-Sentiment-Rating-Predictor-NLP-

streamlit run app.py

pip install -r requirements.txt

📄 License

This project is licensed under the MIT License.

📝 Author

👤 Amir Sohail

---

End-to-End Flipkart product review rating predictor powered by NLP & Extra Trees Classifier (83.2% test accuracy). Features custom text preprocessing, TensorFlow Tokenizer sequence padding, and an interactive Streamlit UI designed to replace unstable ANN training architectures.