In this project, we aimed to enhance the prediction accuracy of mortality rates using two distinct but advanced machine learning approaches: Stacking Ensemble Analysis and Survival Analysis.
Key Highlights:
- Stacking Ensemble Analysis: Combined multiple machine learning algorithms, including Logistic Regression, Random Forest, and Principal Component Analysis, into a single model to improve predictive accuracy. Our ensemble model outperformed individual base models, indicating the effectiveness of combining different algorithmic strategies.
- Survival Analysis: Applied statistical models, such as the Cox Proportional Hazards Model and Random Survival Forests, to understand the likelihood of mortality over time. While the predictive power was modest, these models provided insights into important socioeconomic factors affecting mortality rates.
- Performance Metrics: Utilized negative mean squared error and the concordance index (c-index) to rigorously evaluate model performance.
- Future Direction: Plans to refine these models further by fine-tuning algorithmic parameters and incorporating additional variables to improve accuracy.
The project showcased the application of ensemble methods and survival analysis techniques for complex predictive tasks, providing a comprehensive view of factors that contribute to mortality.
Skills Learned and Improved Through Project
Data Preprocessing:
- Data Cleaning: Achieved proficiency in cleaning large datasets, including handling missing values and outliers.
- Feature Engineering: Sharpened the skill of deriving new, meaningful variables to better train machine learning models.
Machine Learning Algorithms:
- Ensemble Methods: Developed expertise in using stacking ensemble methods to combine different base models for improved predictive accuracy.
Statistical Analysis:
- Survival Analysis: Gained mastery in advanced statistical techniques, including Cox Proportional Hazards and Random Survival Forests, to deal with censored, time-to-event data.
Model Evaluation:
- Cross-Validation: Enhanced skills in implementing K-fold and repeated K-fold cross-validation techniques for robust model evaluation.
- Performance Metrics: Advanced in the use of metrics like Negative Mean Squared Error and Concordance Index (c-index) for model assessment.
Programming and Tools:
- Python and Jupyter Notebook: Refined programming skills in Python and documentation capabilities in Jupyter Notebook, focusing on reproducibility and clarity.
Full Project Write-up
If you want to know more about the project, you can check out the complete write-up below.