Employing High-Performance PETSc Network Simulation for Business Profit Analysis

Business entities such as shops can be modeled as nodes in a traffic network, where vehicular flow at nearby intersections influences a shop's profit. To enable profit-based network analysis, this work presents a methodology for integrating geo-located business and traffic data into high-performance network simulations using the Portable Extensible Toolkit for Scientific Computation (PETSc) and its Data Management Network (DMNetwork) libraries, developed by Argonne National Laboratory (ANL) and UChicago Argonne. Since PETSc and DMNetwork do not natively support mapping business entities to network nodes, we create a new mapping strategy based on geographic coordinates (longitude, latitude). The road network is first constructed using OpenStreetMap (OSM) data and Simulation of Urban Mobility (SUMO) tool. Then, shop locations are obtained via the Google Maps Platform (GMP) and LocationIQ (LIQ) application programming interfaces (APIs). Each shop is assigned to its nearest traffic intersection by calculating the shortest geographic distance, effectively linking business data to the underlying traffic network. We also develop a methodology to obtain real-time traffic data and predict future traffic data at the network nodes for profit analysis. While OSM provides the base topology, GMP and LIQ supply detailed business location information, where LIQ proved to be more comprehensive. The enriched datasets are then used for profit analysis in PETSc/DMNetwork simulations. Simulation results show that utilizing eight processors instead of one reduces computation time by more than 73%, demonstrating the benefit of high-performance PETSc simulation for business analysis.