Predicting Chronic Kidney Disease (CKD)

Models to predict the onset of chronic kidney disease (CKD)

What are included

• Raw data consist of 300 patients, with longitudinal medical record up to 1429 days from baseline
• “Variable explanation.pdf” contains all information about the raw data
• “Predicting chronic kidney disease.pdf” presents the methods and results
• “CSV” folder contain all the raw data and processed data
• “figures” folder contain all the ROC curve
• “models” folder contain all the optimized models
• “predictions” folder contain all the prediction result for model comparison

How to use the notebooks

• exploreData.ipynb is to look through distribution of values
• Run prepareData.ipynb to prepare data for modelling

“Raw” values

• Run Modelling_daybin.ipynb for prediction results of the 180-day bin data
• Run Modelling_agg.ipynb for prediction results of the Aggregated data
• Run Modelling_temporal.ipynb for prediction results of the Temporal data

Categorized (including categorization of drugs by treatment)

• Run Modelling_daybin_cat.ipynb for prediction results of the 180-day bin data
• Run Modelling_agg_cat.ipynb for prediction results of the Aggregated data
• Run Modelling_temporal_cat.ipynb for prediction results of the Temporal data

GFR

• Run Modelling_gfr.ipynb for prediction results of the GFR data