Application of Long Short Term Memory RNN for Minimarket Sales Transaction Prediction

Abstract

This study aims to determine whether it can build a prediction of sales of goods at the Lapan-Lapan Mart by using the Long Short Term Memory Recurrent Neural Network method that can be used to predict the sale of goods. In this study, the data was taken from the Lapan-Lapan Mart, together with data on 10 different items sold every day. The data is then compiled for the level of sales to be weekly and a total of 52 data is obtained for each item so that the total data is amounted to 520. To get the weight in the LSTM calculation, there are two processes, namely forward and backward . the weight will be used to make predictions using the basic formula of the LSTM.Based on the research that has been done, it is known that the highest accuracy of using MAD (Mean Absolute Deviation) is 91 gr (11.61803507) indomie goods and 1.8kg of lemon daia (2.077000464) for the lowest MAD