Kaggle Competition Solutions

A collection of machine learning solutions for various Kaggle competitions and datasets.

📊 Projects

1. Titanic - Machine Learning from Disaster

• Score: 77.99% accuracy
• Model: Random Forest with feature engineering
• Features: Title extraction, family size analysis

2. ConnectX

• Score: 722.2
• Rank: #33
• Approach: Minimax with alpha-beta pruning, 8-10 ply depth

3. House Prices - Advanced Regression

• Target: Top 15%
• Model: 8-model ensemble
• Features: 200+ engineered features

4. Healthcare ML Models

Collection of predictive models for healthcare datasets:

• Heart Disease Prediction (93.48% ROC-AUC)
• Stroke Prediction (78.85% ROC-AUC)
• Diabetes Prediction (82.15% ROC-AUC)
• Sepsis Early Warning (100% ROC-AUC on synthetic data)
• Chronic Kidney Disease (100% ROC-AUC)

🛠️ Tech Stack

• Python 3.8+
• scikit-learn
• pandas, numpy
• imbalanced-learn

📁 Repository Structure

kaggle-competitions/
├── titanic/
├── connectx/
├── house-prices/
├── heart-failure-prediction/
├── stroke-prediction/
├── diabetes-prediction/
├── sepsis-prediction/
└── chronic-kidney-disease/

🚀 Getting Started

# Clone repository
git clone https://github.com/aihearticu/kaggle-competitions.git

# Install dependencies
pip install -r requirements.txt

# Run any model
cd titanic
python3 titanic_improved.py

📈 Results

Competition	Score	Metric
Titanic	77.99%	Accuracy
ConnectX	722.2	ELO Rating
Heart Disease	93.48%	ROC-AUC
Stroke	78.85%	ROC-AUC
Diabetes	82.15%	ROC-AUC
CKD	100%	ROC-AUC

📝 License

MIT