📬 Spam Filter using Probabilistic Classification

A Python-based spam filter that classifies SMS messages as spam or not spam using probabilistic reasoning, word frequency analysis, and text preprocessing. The model is trained directly from a labeled CSV file using basic conditional probability logic (similar to Naive Bayes).

---

📌 Overview

This project demonstrates a custom-built spam classifier trained from real SMS data. It processes a dataset, builds frequency dictionaries for spam and non-spam messages, and uses those statistics to classify new input messages using fundamental probabilistic principles without relying on external machine learning libraries.

Key Features:

• Custom probabilistic classification algorithm
• Text preprocessing and tokenization
• Word frequency analysis
• Real-time message classification
• Educational implementation of Naive Bayes concepts

---

📁 Project Structure

Spam-Filter/
├── 📄 Spam_filter.py     # Main script for building the model and classifying messages
├── 📄 SMS_list.csv       # Input dataset (tab-separated) with labeled SMS messages
└── 📄 README.md          # This documentation file

---

📊 Dataset Requirements

CSV Format: SMS_list.csv

The dataset must be a tab-separated .csv file with two columns: Label and Message

Example format:

spam	WINNER!! You have won a free ticket to Bahamas!
notspam	Are we still on for lunch today?
spam	Congratulations! You've won £1000 cash prize!
notspam	Can you pick up some groceries on your way home?

Important Notes:

• Use spam or notspam as labels
• Ensure no header row in the CSV file
• Use tab separation between label and message
• Place this file in the same directory as Spam_filter.py

---

🧹 Features & Workflow

🔤 Text Preprocessing

• Cleaning: Removes digits, punctuation, and special symbols using regex
• Normalization: Converts all text to uppercase for consistency
• Tokenization: Splits messages into individual words
• Stopword Removal: Filters out high-frequency, non-informative words

🎯 Training Process

• Dictionary Building: Creates two frequency dictionaries (spam_word and n_spam_word)
• Word Counting: Tracks occurrence of each word in spam vs. non-spam messages
• Probability Calculation: Computes conditional probabilities for classification

🔍 Classification Algorithm

• Input Processing: Cleans and tokenizes new messages using the same preprocessing
• Probability Computation: Calculates likelihood of each word belonging to spam/non-spam
• Decision Making: Uses probabilistic reasoning to classify the entire message

---

🚀 Getting Started

Prerequisites

• Python 3.6+ (no additional libraries required)
• Dataset: Properly formatted SMS_list.csv file

Installation & Usage

1. Clone the repository:

   git clone https://github.com/AlphaPruned/Spam-Filter.git
   cd Spam-Filter

2. Prepare your dataset:

   • Ensure SMS_list.csv is in the same directory as Spam_filter.py
   • Verify the tab-separated format with no headers

3. Run the spam filter:

   python Spam_filter.py

4. Test with messages:

   Enter the message: You've been selected to win a free phone!
   m is spam
   

---

🔍 Example Usage

Sample Interactions:

Spam Detection:

Enter the message: WINNER! You have won a lottery of $1000!
m is spam

Non-Spam Detection:

Enter the message: Your package has arrived. Collect it today.
m is not a spam

Marketing Spam:

Enter the message: FREE! Click here to claim your reward now!
m is spam

Normal Conversation:

Enter the message: Let's meet for coffee tomorrow at 3 PM
m is not a spam

---

⚙️ Algorithm Details

Probabilistic Classification Process:

1. Training Phase:

   • Parse labeled dataset
   • Build word frequency dictionaries for each class
   • Calculate prior probabilities for spam/non-spam

2. Prediction Phase:

   • Preprocess input message
   • For each word, calculate: P(word|spam) and P(word|not_spam)
   • Apply Bayes' theorem to determine final classification

3. Decision Rule:

   If P(spam|message) > P(not_spam|message):
       Classify as SPAM
   Else:
       Classify as NOT SPAM
   

---

🔬 Implementation Notes

Strengths:

• Educational Value: Clear implementation of probabilistic classification concepts
• No Dependencies: Uses only Python standard library
• Customizable: Easy to modify preprocessing and classification logic
• Transparent: All probability calculations are explicit and traceable

Limitations:

• No Smoothing: May face zero probability issues with unseen words
• Underflow Risk: Probabilities can underflow for very long messages
• Simple Preprocessing: Basic text cleaning without advanced NLP techniques
• No Cross-Validation: Limited evaluation methodology

Potential Improvements:

• Add Laplace smoothing for unseen words
• Implement logarithmic probabilities to prevent underflow
• Include more sophisticated text preprocessing
• Add model evaluation metrics (precision, recall, F1-score)

---

📈 Performance Considerations

• Speed: Fast classification due to simple probability calculations
• Memory: Efficient storage using Python dictionaries
• Scalability: Performance depends on vocabulary size and message length
• Accuracy: Depends heavily on training data quality and size

---

👤 Author

Arnav Rajesh Kadu

• GitHub: @AlphaPruned

---

🤝 Contributing

Contributions are welcome! Here are some ways you can help:

1. Bug Fixes: Report and fix any issues
2. Feature Enhancements: Add smoothing, logging, or evaluation metrics
3. Documentation: Improve code comments and documentation
4. Dataset: Contribute additional training data

How to Contribute:

1. Fork the repository
2. Create a feature branch (git checkout -b feature/improvement)
3. Commit your changes (git commit -am 'Add improvement')
4. Push to the branch (git push origin feature/improvement)
5. Open a Pull Request

---

📬 Contact

For questions, suggestions, or collaboration opportunities:

• Open an issue on GitHub
• Contact through GitHub profile

---

🙏 Acknowledgments

• Thanks to the SMS Spam Collection dataset contributors
• Inspired by classical machine learning approaches to text classification