Prediksi Indikator Perawatan Menggunakan Random Forest Classification dan Classification and Regression Trees

Abstract

Electronic medical record is a hospital information system facility that contains patient social and medical data to facilitate patient data documentation, improve health services, and access patient data. Medical records can be used to predict length of stay and estimated costs during treatment, so that patients can choose the desired treatment. The method for predicting length of stay uses the Random Forest Classification and Classification and Regression Trees (CART) algorithm which describes the relationship between the dependent and independent variables by producing a binary tree as a decision in the prediction. The model is built by changing the variables into multiclass and binary classes and using hold out split data and cross validation to measure model performance. Each model is compared and the model with the highest accuracy is selected as the best model. The highest model accuracy in the Random Forest Classification is 61.70% and the CART algorithm is 63%. The CART algorithm is the best model for predicting length of stay in this study because it produces the highest accuracy of 63%. Based on the prediction of length of stay, the estimated total cost of hospitalization can be calculated by considering the cost of drugs and inpatient rooms.