Random Acts of Pizza Classification

A machine learning project to predict which Reddit requests for free pizza result in a positive or negative response.

📋 Project Overview

This project uses natural language processing and machine learning to analyze and classify pizza requests from the Reddit community r/RandomActsOfPizza. The goal is to predict whether a request will receive a pizza based on various features including text content, user history, and temporal patterns.

🆕 Modern Updates (2025)

This codebase has been modernized with:

• Updated Dependencies: All libraries upgraded to Python 3.11+ compatible versions
• Modern Python Practices: Type hints, pathlib, f-strings, and PEP 8 compliance
• Refactored Code: Object-oriented design with reusable components
• Enhanced Models: Added gradient boosting and modern ensemble methods
• Improved Feature Engineering: Automated feature extraction pipelines
• Better Documentation: Comprehensive docstrings and usage examples

Key Improvements

1. Deprecated Import Updates

• ✅ pandas.io.json.json_normalize → pd.json_normalize
• ✅ sklearn.grid_search → sklearn.model_selection
• ✅ sklearn.cross_validation → sklearn.model_selection

2. Modern ML Methods

• Added Gradient Boosting classifiers
• Improved feature engineering pipeline
• Better cross-validation strategies
• Modern sentiment analysis with VADER

3. Code Quality

• Removed hardcoded paths - now uses pathlib
• Added comprehensive type hints
• Refactored repetitive code into reusable classes
• Improved variable naming and code organization

🚀 Quick Start

Installation

# Clone the repository
git clone https://github.com/yourusername/Classification-Pizza.git
cd Classification-Pizza

# Create virtual environment
python3 -m venv venv
source venv/bin/activate  # On Windows: venv\Scripts\activate

# Install dependencies
pip install -r requirements.txt

# Download NLTK data (required for text processing)
python -c "import nltk; nltk.download('wordnet'); nltk.download('omw-1.4')"

Basic Usage

Using the Modern Python Module

from pizza_classifier import (
    PizzaDataLoader,
    FeatureEngineering,
    TextPreprocessor,
    PizzaClassifier
)

# Load data
loader = PizzaDataLoader()
train_df, dev_df, test_df = loader.load_data()

# Feature engineering
fe = FeatureEngineering()
train_df = fe.extract_temporal_features(train_df)
train_df = fe.extract_sentiment_features(train_df)
train_df = fe.extract_user_features(train_df)

# Create text corpus
preprocessor = TextPreprocessor()
train_corpus = preprocessor.create_text_corpus(train_df)

# Train classifier
classifier = PizzaClassifier(model_type='random_forest')
classifier.fit(
    train_corpus,
    fe.create_feature_matrix(train_df),
    train_df['requester_received_pizza'].values
)

# Make predictions
predictions = classifier.predict(dev_corpus, dev_features)

Using Legacy Notebooks

The original Jupyter notebooks are preserved in the repository:

• Clean_Notebook_Compiled.ipynb - Main analysis notebook
• W207_Final_Project_Baseline_v4.ipynb - Baseline models
• Learning_Notebook_Compiled.ipynb - Experimental approaches

📊 Results

Our best models achieve:

• Random Forest: 83% accuracy, F1-score 0.81, AUC 0.71
• Logistic Regression: 74% accuracy, AUC 0.73
• Gradient Boosting: 80%+ accuracy with proper tuning

Key findings:

• Temporal features (hour, day) are strong predictors
• Text features with bigrams perform well
• Sentiment analysis provides marginal improvement
• Simple models (Logistic Regression) are competitive with complex ensembles

🔍 Code Review Findings

Issues in Original Code

1. Deprecated Libraries (Fixed ✅)

   • Old sklearn imports
   • Outdated pandas JSON normalization
   • Legacy string formatting

2. Code Duplication (Fixed ✅)

   • Repetitive data preprocessing
   • Duplicate model evaluation code
   • Multiple similar notebooks

3. Hardcoded Values (Fixed ✅)

   • File paths
   • Magic numbers
   • Configuration scattered throughout

4. Missing Best Practices (Fixed ✅)

   • No type hints
   • Limited error handling
   • No virtual environment specification
   • Inconsistent naming conventions

📁 Project Structure

Classification-Pizza/
├── data/
│   ├── train.json          # Training data
│   └── test.json           # Test data
├── pizza_classifier.py     # ✨ NEW: Modern implementation
├── requirements.txt        # ✨ NEW: Dependency management
├── README.md              # ✨ UPDATED: This file
├── Clean_Notebook_Compiled.ipynb
├── W207_Final_Project_Baseline_v4.ipynb
└── [Other legacy notebooks...]

🛠️ Technology Stack

• Python: 3.11+
• ML Frameworks: scikit-learn, XGBoost, LightGBM
• NLP: NLTK, VADER Sentiment
• Data: pandas, numpy
• Visualization: matplotlib, seaborn

📝 Features

Text Features

• TF-IDF / Count vectorization
• N-gram analysis (unigrams, bigrams, trigrams)
• Sentiment scores (VADER)
• Text preprocessing and cleaning

Temporal Features

• Hour of day
• Day of week
• Month/seasonality
• Weekend indicator

User Features

• Account age
• Reddit karma (upvotes/downvotes)
• Comment/post ratios
• Subreddit activity
• Previous pizza requests

🎯 Models Implemented

• Logistic Regression (L1/L2 regularization)
• Random Forest
• Naive Bayes
• Support Vector Machines
• Gradient Boosting
• AdaBoost
• Ensemble methods

📚 Data

Data is from the Kaggle competition: Random Acts of Pizza

The dataset contains:

• 4,040 training samples
• 1,630 test samples
• 32 features per sample
• Binary classification target

Data is located in the /data directory.

🤝 Contributing

This is an educational project. Feel free to fork and experiment!

📄 License

MIT License - see LICENSE file for details

👥 Original Authors

W207 Final Project Team:

• Erika Lawrence
• Leslie Teo
• Jen Jen Chen
• Geoff Stirling

🔄 Modernization

Code modernization and updates: 2025