Spam Detection using RNN

Detecting Spam with Deep Learning

Project Overview

In this project, I explore the application of Recurrent Neural Networks (RNN) in spam detection. My model is trained on a dataset of labeled messages(5572 sample) and achieves a high accuracy in distinguishing between spam and legal messages.

Key Features

• Deep Learning: Utilizes RNN to learn patterns and sequences in data
• High Accuracy: Achieves a high accuracy in spam detection(99.3% training accuracy and 95.52% testing accuracy)
• Open-Source: Fully open-source and available for modification and improvement

Dataset

• Description: [Dataset has 2 columns Text and Labels(spam or not-spam)]
• Source: [https://archive.ics.uci.edu/dataset/228/sms+spam+collection]

Data Preprocessing

• Lowercasing
• Punctuation Removal
• Stop Words Removal
• Url Handled
• Whitespaces Removal
• Tokenization
• Lemmatization

Model Architecture

• Embedding: [Ivocab size=9011, max words=50 , 32]
• RNN: [SimpleRNN with 32 units]
• Activation Functions: [relu]
• Optimizer: [Adam]
• Metrics: [Accuracy]
• Dropout Layer: [After embedding(0.3) and after SimpleRNN(0.5)]

Results

• Accuracy: [Training Accuracy:99.3% and Testing Accuracy:95.52%]
• Loss: [Training Loss:0.02 and Testing Loss:0.19]

Usage

1. Clone the repository: git clone https://github.com/Debopam-Pritam2014/Spam-Detection-Using-RNN.git

Contributing

Contributions are welcome! Feel free to fork the repository, make changes, and submit a pull request.

License

This project is licensed under the MIT License.