Publication

Abstract : Estimating “Recovered length of stay” is useful in the better utilization of resources, effective planning treatment and cost-effectiveness in managing acute malnutrition patients. This study proposes a machine learning approach to improve the precision in estimating “Recovered length of stay” for patients suffering from acute malnutrition. The present work analyzes the seven Machine Learning (ML) models to estimate the recovered length of stay. The best two models were chosen based on the standard evaluation metrics such as MSE (Mean Squared Error), Coefficient of determination (R2), MAE (Mean Absolute Error), MAPE (Mean Absolute Percentage Error). Further, the input features to the best two models are reduced using one of the commonly used dimensionality reduction algorithms known as Principal Component analysis (PCA). The proposed approach is fully tested using the particular dataset for acute malnutrition cases and estimates the recovered periods for the patients. The results of the experimental tests show that the use of multiple regression models, especially Decision Tree (DT) and Support Vector Regression (SVR), results in the greatest improvements in the estimation of in case of acute malnutrition. In addition, the models presented a R-squared value of 0.9459 and 0.9316, and they are the best models to capture the complex features of patient recovery times.