Pengujian Long-Short Term Memory (LSTM) Pada Prediksi Trafik Lalu Lintas Menggunakan Multi Server

Abstract

This study presents a test of the long short term memory (LSTM) algorithm on traffic prediction with multi edge server and cloud server architectures. IoT sensors located on the roadside such as cameras and location data on each driver are used and stored in the data center. When a driver sends a travel time request to a nearby edge server, traffic predictions will be made on the edge server or cloud server. Server selection is made based on the destination location of the driver's request. If the destination is in the edge server area, traffic predictions are made on the edge server. However, if the destination is in the cloud server area, traffic predictions are made on the cloud server. Then to predict traffic traffic is done with LSTM. following modeling is made with a density of 128 and a density of 256. By learning from previous traffic, LSTM with a greater density gets a proportion of errors, namely RMSE 10.78%, MAE 8.24%, and MAPE 19.87%.