Penerapan Metode Backpropagation Dalam Memprediksi Jumlah Penjualan Oli Shell

Abstract

The application of a prediction is very important to do in research, so that research becomes faster and directed. Just as in predicting the number of shell oil sales, studies and the use of appropriate methods are needed to obtain optimal results. The data used in this study is sales data from PT. Mitra Petra Sejahtera Kota Medan from 2012 to 2017. The algorithm used to make this prediction is the backpropagation algorithm. This algorithm is used to predict future results based on previous data. There are 6 architectural models used in the backpropagation algorithm, among others, 4-2-1 which will later produce predictions with 83% accuracy, 4-3-1 = 78%, 4-4-1 = 83%, 4- 5-1 = 78%, 4-8-1 = 100% and 4-10-1 = 72%. The best architecture of these 6 models is 4-8-1 with an accuracy rate of 94% with a level of Error 0.001, MSE = 0.04133616. so this architectural model is good enough to be used to predict the amount of shell oil sales.