A machine learning project that analyzes customer reviews from the Amazon Alexa dataset to predict sentiment (positive or negative). The project leverages a pipeline approach with TF-IDF vectorization and Logistic Regression for efficient preprocessing and classification.
- Preprocessing of textual review data.
- Sentiment classification using a scikit-learn pipeline.
- TF-IDF feature extraction.
- Logistic Regression model for sentiment prediction.
- Evaluation metrics: accuracy, confusion matrix, classification report.
- Jupyter Notebook for training and visualization.
- Optional Streamlit web app (if applicable in future).
The dataset used is the Amazon Alexa Reviews dataset, which contains:
verified_reviews: Customer review textfeedback: Target variable (1 = positive, 0 = negative)
- Python
- scikit-learn
- pandas, numpy
- matplotlib, seaborn (for visualization)
- TF-IDF Vectorizer
- Logistic Regression
-
Clone the Repository
git clone https://github.com/itz-Pratham/Sentiment_Analysis_Amazon.git cd Sentiment_Analysis_Amazon -
Install Dependencies
pip install -r requirements.txt
-
Run the Jupyter Notebook
jupyter notebook Sentiment_Analysis.ipynb
- Achieved high accuracy with Logistic Regression and TF-IDF features.
- Clear distinction between positive and negative reviews observed in classification results.
(You can include a confusion matrix or accuracy figure here)
- Integrate with a Streamlit web app for user interaction.
- Add support for more ML models (e.g., Naive Bayes, SVM).
- Implement hyperparameter tuning using GridSearchCV.
- Dataset from Kaggle
- Inspired by real-world sentiment analysis applications.
Created with 💻 by Pratham Feel free to connect or suggest improvements!
Would you like help generating a badge section or requirements.txt as well?