Analysis of Backpropagation Method in Predicting Drug Stock

Abstract

Health is one of the needs of the society to be able to carry out activities in daily life. The high public need for health makes the provision of drug stocks being an important aspect for every health service, one of which is the Public Health. To find out how many drugs are received, the Public Health conducts a data collection on the drug stock that is carried out. However, the Public Health could not determine the exact amount of drug stock needed so that some drugs expired due to excess drug stock. With these problems, an analysis using the Backpropagation method was carried out to predict how much drug stock was needed. The purpose of this study is to ensure that the stock of drugs that enter the Public Health is in accordance with the needs of the Public Health so that it can reduce the expired of the drugs. For analysis using the Backpropagation Method, there are two processes used, namely: with the training process to find new weights which will later be used in the second process, namely the testing process. testing is only carried out on the forward propagation process. To carry out the training process and testing process, data from 2019-2020 that has been recapitulated will be used. The results of the analysis using the Backpropagation method show that the highest accuracy results are 88.0356% at epoch 900, learning rate is 0.001, and goal is 0.00001 with the lowest MAPE (Mean Absolute Percentage Error) is 11.964% which shows that the ability of forecasting or analysis models good because how much in the range of 10-20%