Project Overview

The project aims to predict pregnancy success using advanced machine learning techniques developed for the LG Aimers Hackathon.

Final Performance

Private Score: 0.74159 (TOP 22%)

👥 Team Members

김재원	신동우	홍성인

Technology Stack

• Python 3.9.7
• pandas 1.3.5
• numpy 1.21.6
• matplotlib 3.5.1
• seaborn 0.11.2
• scikit-learn 1.0.2
• imbalanced-learn 0.8.1
• catboost 1.0.6
• scipy 1.7.3

Key Features

Data Preprocessing

• Class-based missing value handling
• Advanced feature engineering
• Strict data leakage prevention

Modeling Strategy

• 7-fold cross-validation
• CatBoost algorithm
• Ensemble techniques

Performance Optimization

• Detailed hyperparameter tuning
• Class imbalance handling
• Feature importance-based selection

Performance Evaluation

• Evaluated using ROC AUC score
• Multiple submission strategies:
  1. Single best-performing model
  2. Average ensemble of all folds
  3. Weighted ensemble of top 3 folds
  4. Weighted ensemble of top 5 folds

Solution Highlights

• Advanced machine learning techniques
• Sophisticated feature selection
• Robust cross-validation methodology