Pokémon Legendary Classification

A machine learning project that predicts whether a Pokémon is Legendary based on its base stats. This project demonstrates data preprocessing, feature engineering, and binary classification using Logistic Regression, achieving 94.17% accuracy.

Project Overview

This project analyzes Pokémon statistics from Generations 1-6 to build a predictive model that classifies Pokémon as Legendary or non-Legendary. The model uses various features including base stats (HP, Attack, Defense, etc.), type information, and engineered features to make predictions.

Key Features:

• Binary classification task (Legendary vs. Non-Legendary)
• Comprehensive data preprocessing and feature engineering
• Feature analysis with correlation matrices
• Logistic Regression model with 94%+ accuracy
• Exploratory data analysis with visualizations

Dataset

The dataset contains information about 800 Pokémon from Generations 1-6, including:

Features:

• #: Pokémon ID number
• Name: Pokémon name
• Type 1/Type 2: Pokémon types (e.g., Grass, Poison, Fire)
• Total: Sum of all base stats
• HP: Hit Points
• Attack: Physical attack power
• Defense: Physical defense
• Sp. Atk: Special attack power
• Sp. Def: Special defense
• Speed: Speed statistic
• Generation: Generation number (1-6)
• Legendary: Target variable (True/False)

Engineered Features:

• Dual Type: Indicator if Pokémon has two different types
• Mega: Indicator if Pokémon is a Mega Evolution

Source: The dataset is included in the repository as Pokemon.csv

Technologies Used

• Python 3.x
• Libraries:
  • pandas - Data manipulation and analysis
  • numpy - Numerical computations
  • scikit-learn - Machine learning models and preprocessing
  • matplotlib - Data visualization
  • seaborn - Statistical data visualization

Installation

Prerequisites

• Python 3.7 or higher
• pip package manager

Setup Instructions

1. Clone the repository:

   git clone https://github.com/yourusername/pokemon-classification.git
   cd pokemon-classification

2. Create a virtual environment (recommended):

   python -m venv venv
   
   # On Windows
   venv\Scripts\activate
   
   # On macOS/Linux
   source venv/bin/activate

3. Install required packages:

   pip install pandas numpy scikit-learn matplotlib seaborn tensorflow

   Or install from requirements file (if available):

   pip install -r requirements.txt

4. Launch Jupyter Notebook:

   jupyter notebook

5. Open PokemonClassification.ipynb and run all cells

Usage

1. Open the Jupyter notebook PokemonClassification.ipynb
2. Run cells sequentially from top to bottom
3. The notebook will:
   • Load and explore the Pokémon dataset
   • Perform data preprocessing and feature engineering
   • Train a Logistic Regression model
   • Display model performance metrics
   • Show visualizations and insights

Expected Output:

• Model accuracy: ~94.17%
• Correlation heatmap of features
• Feature importance analysis
• Predictions on test data

Results

Model Performance

• Algorithm: Logistic Regression
• Accuracy: 94.17%
• Train/Test Split: 70/30 (stratified)
• Scaling: MinMaxScaler for feature normalization

Key Insights

1. Strong Predictors:

   • Total base stats show the strongest correlation with Legendary status
   • Individual stats (HP, Attack, Defense, Sp. Atk, Sp. Def, Speed) all contribute positively

2. Feature Correlations:

   • High multicollinearity exists between Total stats and individual stats
   • Attack and Sp. Atk show positive correlation
   • Defense and Sp. Def are moderately correlated

3. Type Analysis:

   • Both Type 1 and Type 2 show weak correlation with Legendary status
   • Dual-type Pokémon are slightly more likely to be Legendary

4. Mega Evolution:

   • Mega Pokémon show strong correlation with Legendary classification
   • Only 65 Legendary Pokémon in the dataset (8.1%)

Sample Prediction

The model correctly predicts high-stat Pokémon as Legendary. For example, a Pokémon with stats:

• Type: Dragon/Psychic
• Total: 700
• HP: 80, Attack: 130, Defense: 100
• Sp. Atk: 160, Sp. Def: 120, Speed: 110
• Generation: 3

Prediction: 35.4% probability of being Legendary (threshold: 50%)

Project Structure

pokemon-classification/
│
├── PokemonClassification.ipynb   # Main Jupyter notebook
├── Pokemon.csv                    # Dataset (in sample_data/ if using Colab)
├── README.md                      # This file
└── requirements.txt              # Python dependencies (optional)

Notes

• The notebook was originally created in Google Colab (note the sample_data/ path)
• The dataset includes Mega Evolutions but excludes Meganium and Yanmega when identifying Megas
• Missing Type 2 values are filled with Type 1 (single-type Pokémon)
• The Legendary column is converted to binary (0/1) for modeling

Contributing

This is a personal portfolio project, but suggestions and feedback are welcome! Feel free to:

• Open an issue for bugs or suggestions
• Fork the repository and submit pull requests
• Share your own improvements or variations

License

This project is available under the MIT License. Feel free to use it for learning and portfolio purposes.

Acknowledgments

• Dataset source: Kaggle Pokémon Dataset
• Pokémon and all related properties are © Nintendo, Game Freak, and The Pokémon Company
• Created as a machine learning portfolio project