Predicting Hospital Readmission for Diabetic Patients
Using patient data from a hospital stay, can we predict whether that patient would be readmitted?
Description
The need for repeated episodes of hospital care disrupts the patient’s life, costs the healthcare industry billions of dollars each year, places demand on hospital bed capacity, and threatens the viability of hospitals with high readmission rates in the form of Medicare linking payment to the quality of hospital care. In this project, we attempt to predict whether or not a diabetic patient would be readmitted to the hospital using demographic and health related information gathered in the initial hospital admission.
Data
Diabetes 130-US hospitals for years 1999-2008 Data Set from the UCI Machine Learning Repository.
Contents
Data Exploration folder - All data exploration code and results
dataset_diabetes - Datasets used
Cost Matrix.xlsx - Explanation of the cost function created for model evaluation
Initial Modeling.ipynb - Modeling iteration 1
Updated Modeling.ipynb - Modeling iteration 2
Random Forest using H2O.ipynb - Modeling iteration 3
Cost Sensitive Random Forest using CostCla.ipynb - Modeling iteration 4
Technical Report and Executive Summary.pdf
Presentation Deck.pdf - Video Recording
Tools
- Python
- Sklearn
- Seaborn
- Matplotlib
- CostCla
- H2O
Authors
Samuel Sears @ssears219
Andrea Fox @anfox86
Jolene Branch