Sentiment Analysis on IMDB Movie Reviews

Project Overview

This project focuses on performing sentiment analysis on IMDB movie reviews using deep learning techniques with an LSTM (Long Short-Term Memory) model. The dataset used contains 50,000 movie reviews labeled as positive or negative.

Technologies Used

• Python
• TensorFlow/Keras
• Pandas
• Scikit-learn
• Kaggle API

Dataset

The dataset used in this project is the IMDB Dataset of 50K Movie Reviews, which was downloaded from Kaggle.

Steps Involved

1. Dataset Downloading:

   • The dataset is downloaded using the Kaggle API.
   • The zip file is extracted to access the CSV file.

2. Data Preprocessing:

   • The dataset is loaded using Pandas.
   • Sentiment labels are converted to numerical values (positive: 1, negative: 0).

3. Train-Test Splitting:

   • The dataset is split into training (80%) and testing (20%) sets.

4. Tokenization and Padding:

   • Tokenization is applied to convert text to sequences.
   • Padding is used to ensure uniform sequence length.

5. LSTM Model Building:

   • An LSTM model is created with the following layers:
     • Embedding layer
     • LSTM layer
     • Dense output layer with a sigmoid activation function
   • Model is compiled using binary cross-entropy loss and the Adam optimizer.

6. Model Training:

   • The model is trained with 5 epochs and a batch size of 64.
   • Validation split of 20% is used.

7. Model Evaluation:

   • The model is evaluated on the test data.
   • Accuracy and loss metrics are reported.

8. Prediction Function:

   • A function is implemented to predict sentiment based on user input reviews.

Results

• The model achieved satisfactory accuracy on the test set.
• Example predictions:
  • "This movie was not so interesting." -> Negative
  • "This movie was very amazing." -> Positive

How to Run the Project

1. Clone the repository from GitHub.
2. Install required dependencies using:

   pip install -r requirements.txt

3. Run the Jupyter Notebook to train and evaluate the model.

Future Improvements

• Increase the dataset size to enhance model performance.
• Tune hyperparameters for better accuracy.
• Experiment with different neural network architectures.

License

This project is under the MIT License.

Author

Arpan Pramanik