https://github.com/vaneeza-7/spam-detection-using-feedforwardnn-and-word2vec

Spam Detection using 2 layer deep feed forward neural network and word2vec embeddings
https://github.com/vaneeza-7/spam-detection-using-feedforwardnn-and-word2vec

Last synced: 8 months ago
JSON representation

Spam Detection using 2 layer deep feed forward neural network and word2vec embeddings

• Host: GitHub
• URL: https://github.com/vaneeza-7/spam-detection-using-feedforwardnn-and-word2vec
• Owner: Vaneeza-7
• License: mit
• Created: 2025-02-17T21:25:18.000Z (8 months ago)
• Default Branch: main
• Last Pushed: 2025-02-17T21:30:49.000Z (8 months ago)
• Last Synced: 2025-02-17T22:27:54.332Z (8 months ago)
• Language: Jupyter Notebook
• Homepage:
• Size: 1.28 MB
• Stars: 0
• Watchers: 1
• Forks: 0
• Open Issues: 0
• Metadata Files:
  • Readme: README.md
  • License: LICENSE

Awesome Lists containing this project

README

          
# **Spam Detection Using a Feed-Forward Neural Network with Word Embedding**  

## **Overview**  

This project implements a **two-layer feed-forward neural network** for **spam detection** using **custom Word2Vec embeddings**. The model classifies text messages as **spam or not spam** based on learned word representations.  

## **Dataset**  

- **Source**: [Spam or Not Spam Dataset](https://www.kaggle.com/datasets/ozlerhakan/spam-or-not-spam-dataset?resource=download)  

- **Task**: Classify messages as **spam (1)** or **not spam (0)**  

## **Tasks**  

### 🛠 **1. Data Preprocessing**  

- Balance the dataset and clean text data for model training.  

### 📖 **2. Model Training**  

#### 🔡 **a. Word2Vec Embeddings**  

- Implement **Word2Vec from scratch** using **logistic regression with negative sampling**.  

- Each word is represented as a **10-dimensional embedding**.  

#### 🧠 **b. Neural Network Architecture**  

- **Input**: 12 words per message (each as a 10-dimensional vector).  

- **Hidden Layer**: 2 nodes, each with a size of 8.  

- **Output Layer**: 1 node for binary classification.  

- **Training**: Implement **forward & backward pass, stochastic gradient descent (SGD)** without using external libraries.  

### 📊 **3. Model Evaluation**  

- Evaluate performance using **accuracy, precision, recall, and F1-score**.  

- Visualize model performance using a **confusion matrix**.  

🚀 **This project demonstrates an end-to-end spam detection pipeline, from word embeddings to deep learning classification (from scratch)!**