Decision-Level Fusion on Healthcare

Post Graduate Program, Department of Mathematics, Universitas Gadjah Mada, Indonesia

^* Corresponding author: astriayunastiti@mail.ugm.ac.id

Abstract

Patient care management is a crucial aspect that affects the quality of life of patients and the operational efficiency of hospitals. Patient care can be broadly categorized into two main categories: inpatient and outpatient. The objective of this study is to develop a machine learning model that can accurately predict whether a patient should be classified as inpatient or outpatient based on their laboratory test results. Five classification methods are applied in this study: Logistic Regression, Support Vector Machine (SVM), Random Forest, and Gradient Boosting. The feature variables in the dataset include laboratory test results and patient demographic information, such as Haematocrit, Haemoglobins, Erythrocyte, Leucocyte, Thrombocyte, MCH, MCHC, MCV, Age, and Gender. The target variable is the type of patient care, coded as 1 for inpatient and 0 for outpatient. This study also implements Decision Fusion to enhance prediction accuracy and stability. After preprocessing to detect and remove outliers, the data is split into training and testing sets. The models are then fitted and tested on the test data. Predictions from the four classification methods are combined using decision level fusion as majority voting, score level and weighted voting to obtain the final prediction. Thus, the fusion method can provide better performance compared to individual models by leveraging the collective strength of all the models used. In this study, we use two scenarios, a false negative ratio of 1% and a false negative ratio of 5% to show the performance of decision fusion.