Application of Spatial Temporal Graph Neural Network in Analyzing the Distribution of Goods Shipping with Dominating Set Technique

Abstract

A logistics service company may face capacity issues due to distribution delays, resulting in goods accumulating in branch offices with unknown locations. To resolve this problem, we will implement the Spatial-Temporal Graph Neural Network (STGNN) combined with the dominating set technique to predict these branch office locations. The STGNN utilizes graph theory to represent relationships between branch offices in Indonesia. Simulation data on goods shipments across Indonesia are observed for 30 days, categorized as spatial-temporal data, and fed into the STGNN. This process involves three stages: node embeddings, training, and testing/forecasting. We implement some Artificial Neural Network (ANN) models with various hidden layer architectures. The results show that the best model of ANN is cascade forward metwork and the MSE 1,6714×〖10〗^(-9).