Prediksi Pasien Pusat Kesehatan Masyarakat Menggunakan Machine Learning

Abstract

The fluctuating nature of patient visits makes it difficult for hospital management to plan, so it is important to predict patient visits by community health centers (PusKesMas) based on gender. The purpose of this study is to predict whether patients who come for treatment at the community health center can be served immediately, the supply/stock of drugs can meet the needs of patients and the availability of sufficient medical equipment, so that community health center services improve for the better. Based on good performance in solving the problems that have been formulated, the methods used are Random Forest (RF) and Extreme Gradient Boosting (XGBoost). The programming language used is Python using Google Colab. The stage of separating tain and test data using the scikit-learn train_test_split module with a percentage of 70% for train data and 30% for test data produces an accuracy in RF of 0.69 while in XGBoost it is 0.93. The results of the confusion matrix from XGBoost are true positive (TP), namely data that is predicted correctly and precisely as much as 53, false negative (FN) worth 3, false positive (FP) worth 2 and 1, true negative (TN) worth 40, 4, 1, 46. Meanwhile, the results of the XGBoost classification report model from the weighted Average precision value of 0.93, the recall value of 0.93 and the F1-Score value is also 0.93. These results indicate that the model used has good quality performance, so it is worthy of use. The application carried out is with the XGBoost data classification to assess patient visits in the next 5 years, with a prediction of achieving 93% accuracy.