Predicting Medicine-Stocks by Using Multilayer Perceptron Backpropagation

Abstract

Artificial neural network has a lot of ability in controlling the error rate to formulate some of its functions as a supervised method. Medicine is one of the major needs for each patient, so that every hospital should know how much inventory of drugs is used and needed by patients every day. This study uses artificial neural network with multilayer perceptron backpropagation as a solution for predicting drug stocks. Prediction of drug stocks using medicine prior period stock data for three years is used to get the predicted results with a small error rate. Backpropagation algorithm using the error output is used to change the weights in the backward direction. To get this error, forward stage should be done first. The results of experiments using backpropagation with the configuration of 0.04 mementum which has training rate of 0.001 gets the value of MSE of 0.00001.