Predicting the amount of digestive enzymes medicine usage with LSTM

Abstract

Medicines are widely used to prevent or cure illness. One of the medicines, which often used to relieve stomach pain, is a medicine that contains digestive enzymes. Hospitals and other health institutions much need this type of medicines. Hospitals and other health institutions should ensure the availability of medications for patients. This situation forces health institutions to deal with the uncertainty of medicine usage. Hospitals as one of the health institutions have some challenges. One of the challenges that must be faced is to ensure the availability of medicines for patients. The ability to predict can help ensure medicines availability in hospital. In this study will presents the forecasting model using Long Term Short Memory (LSTM) method to predict the need for medicines that contain digestive enzymes in the hospital. This method is chosen because it is known to have high accuracy to predict stationary data. One of the methods used in input identification for the LSTM method is by using the Autocorrelation Function (ACF) and Partial Autocorrelation Function (PACF). The results of this study indicate that the use of LSTM method suitable for time series forecasting in the historical dataset, with 12.733 Root Mean Square Error (RMSE) value.