Pokémon Legendary Status Predictor

Project Overview

This project implements a machine learning-based system to predict whether a Pokémon is legendary based on its characteristics and statistics. The system employs a Random Forest Classifier trained on comprehensive Pokémon data, with a focus on handling class imbalance and optimizing model performance.

Technical Implementation

Data Processing Pipeline

1. Data Collection and Preprocessing

   • Utilizes a comprehensive Pokémon dataset containing base stats, physical characteristics, and type information
   • Implements data cleaning and feature engineering
   • Handles missing values using SimpleImputer with most frequent strategy

2. Feature Engineering

   • Numerical features: HP, Attack, Defense, Special Attack, Special Defense, Speed, Height, Weight, Generation
   • Categorical features: Primary and Secondary Types (one-hot encoded)
   • Target variable: Binary classification (Legendary vs Non-Legendary)

3. Model Architecture

   • Random Forest Classifier with optimized hyperparameters
   • Class weights to address imbalanced data
   • SMOTE (Synthetic Minority Over-sampling Technique) for data augmentation to help balance training dataset
   • Stratified sampling for training/test split

System Components

1. Data Analysis Module (pokemon_models.py)

• Implements data structures for Pokémon statistics
• Provides methods for statistical analysis
• Generates visualizations for data exploration

2. Machine Learning Module (pokemon_ml.py)

• Implements the core prediction system
• Handles data preprocessing and feature engineering
• Trains and evaluates the Random Forest model
• Provides model interpretability through feature importance analysis
• Implements prediction interface for new Pokémon

3. Dashboard Interface (dashboard.py)

• Interactive Streamlit-based web interface
• Real-time model evaluation and visualization
• Interactive prediction interface
• Comprehensive performance metrics display

Installation and Setup

Prerequisites

• Python 3.8 or higher
• Required packages (see requirements.txt):
  • pandas: Data manipulation and analysis
  • scikit-learn: Machine learning implementation
  • streamlit: Web interface framework
  • plotly: Interactive visualizations
  • imbalanced-learn: Handling class imbalance

Installation Steps

1. Clone the repository
2. Create and activate a virtual environment:

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

3. Install dependencies:

   pip install -r requirements.txt

Running the Application

streamlit run scripts/dashboard.py

Technical Details

Model Architecture

• Algorithm: Random Forest Classifier
• Hyperparameters:
  • n_estimators: 100
  • max_depth: 10
  • min_samples_split: 10
  • min_samples_leaf: 5
  • max_features: 'sqrt'
  • class_weight: balanced

Performance Optimization

• SMOTE implementation for class balancing
• Stratified sampling for representative test sets
• Feature importance analysis for model interpretability
• Regularization techniques to prevent overfitting

References

• Pokémon Dataset: Publicly available Pokémon statistics database
• Scikit-learn Documentation: Machine learning implementation reference
• Streamlit Documentation: Web interface framework reference
• Imbalanced-learn Documentation: Class imbalance handling reference