A sample study using various data science tools on business case for improving hospital in-patient process, performed for a university hospital dataset
Business Case - Data Science for Hospital and Medical Industry Generic Hospital is looking to improve their in-house patient service, the following notebook is meant to serve as an example of using data analysis and machine learning to improve the in-house patient process. The following are objectives of this analysis
1 .Understand the key relationships which impact the length of stay of patients in hospitals in the provided data. Analysis to be completed include basic statisical analysis, univariate analysis, multivariate analysis. 2. Build a prediction model to predict length of stay in hospital given various symptoms. We will build tree based models as well as deep learning models. The metrics will be R2, MAE, RMSE on target varible of length of stay. Best models will be based on metrics, test data, variation in trees, model explianibility. Understand and research which factor impact the length of stay 3.Perform a clustering analysis using K-means to understand various kinds of patients group. Recommend if we provide tailer-made services for each group 4. Provide any recommendations to improve services for patients based on data science tools.
Models used
- Cluster analysis K-means
- Random Forests
- Partial Dependency Analysis
- LightGBM
- XGboost regressor
- Keras MLP
Optimizations using hyperopt library
Please refer to the attached notebook for detailed analysis
Model deployement is completed using Docker/FastAPI.