Using Machine Learning Method to Predict Treatment Pathway in Systemic Lupus Erythematosus

The UK Renal Registry has set goals to analyze and report data on renal patients who are suffering relative diseases and who are on dialysis. There is one another disease, which is called Systemic Lupus Erythematosus. This is one of rheumatic diseases but always complicated with kidney also occurred in some patients. In order to provide the suitable treatment pathway for each patient, it is necessary to identify the predictors for this kind of disease. In this paper, one machine learning method will be applied to analyze disease factors about SLE, which is called logistic regression, to identify which one is a good predictor or not. Before that, one statistical method was used to check data correlations between each feature and SLE, namely chi-square test, to select those with good significance and filter out those without. With our collected SLE patients’ dataset, the occurrence of missing data also spread around various kinds of variables, to deal with that, we applied MICE to do the imputation of the missing data. By the way, one regularization method, elastic-net, was used to prevent overfitting of logistic regression model.