Readme.md

Sentiment Analysis with LSTM

This project implements a Sentiment Analysis model using LSTM (Long Short-Term Memory) networks on the IMDB movie reviews dataset.

📌 Project Details

• Objective: Classify movie reviews as positive or negative.
• Tech Stack: Python, TensorFlow/Keras, Pandas, NumPy, Matplotlib, Scikit-learn.
• Approach:
  • Preprocess the text data.
  • Tokenize and pad sequences.
  • Train an LSTM neural network to learn sentiment patterns.
  • Evaluate the model performance.

📊 Dataset

• Name: IMDB Dataset of 50K Movie Reviews.
• Source: Kaggle
• Description:
  • 50,000 reviews labeled as positive or negative.
  • Balanced dataset (25,000 positive / 25,000 negative).

📈 Training Results

The model was trained for 5 epochs with the following performance:

• Epoch 1: Accuracy = 51.48%, Loss = 0.6938, Val Accuracy = 54.51%, Val Loss = 0.6889
• Epoch 2: Accuracy = 56.55%, Loss = 0.6736, Val Accuracy = 53.46%, Val Loss = 0.7004
• Epoch 3: Accuracy = 56.17%, Loss = 0.6803, Val Accuracy = 58.44%, Val Loss = 0.6674
• Epoch 4: Accuracy = 59.64%, Loss = 0.6568, Val Accuracy = 60.69%, Val Loss = 0.6646
• Epoch 5: Accuracy = 72.97%, Loss = 0.5580, Val Accuracy = 82.66%, Val Loss = 0.4459 s Final Test Accuracy: 82.66%

⚙️ Workflow

1. Data Loading: Import the dataset into Pandas DataFrame.
2. Preprocessing:
   • Remove HTML tags and special characters.
   • Convert text to lowercase.
   • Map labels: positive → 1, negative → 0.
3. Tokenization & Padding:
   • Convert words into integer sequences using Keras Tokenizer.
   • Pad sequences to ensure each review has the same length (200).
4. Model Building:
   • Use Embedding + LSTM + Dense layers.
5. Training:
   • Optimizer: Adam
   • Loss: Binary Crossentropy
   • Epochs: 5
   • Batch Size: 128
6. Evaluation:
   • Measure accuracy on test set.
   • Plot training vs validation accuracy.
7. Prediction: Test the model with custom reviews.

🏗️ Model Architecture

• Embedding Layer: Input dim = 10,000, Output dim = 64, Input length = 200
• LSTM Layer: 128 units
• Dropout Layer: 0.5
• Dense Layer: 64 units, ReLU activation
• Dropout Layer: 0.3
• Output Layer: 1 unit, Sigmoid activation

▶️ How to Run

1. Clone or download this repository.
2. Download the dataset and place IMDB Dataset.csv in the project folder.
3. Install dependencies:

   pip install -r requirements.txt
   

4. Run the script:

   python main.py
   

🚀 Future Improvements

• Use pre-trained embeddings like GloVe.
• Try BiLSTM/GRU architectures.
• Experiment with transformers (BERT) for better accuracy.
• Deploy the model as a web app using Flask or Streamlit.