Optimization of Seating Arrangement in Sports Competition Venues Based on Recurrent Graph Neural Network

Abstract

Seating in sports venues is a type of seating where rows are much more frequent than in other types of auditorium seating. Traditional seating looks to maximize capacity over fan experience. The stadium experience can be enhanced by recommending appropriate seating by viewer’s preferences and past behavior. Group customers would want to sit together when they go to a sports arena, but sometimes there are not enough seats available together. This study's methodology for improving viewing experience by optimizing seating arrangement using game type, historical ticket sales and opponent team is offered in this study as a way to get around this drawback. The data are collected from Los Angeles Memorial Sports Arena seating dataset. Afterward, the data are fed to pre-processing. In pre-processing segment, the noise is removed and data is enhanced using Adaptive Robust Cubature Kalman Filter (ARCKF). The outcome from the pre-processing data is transferred to the Recurrent Graph Neural Network (RecGNN).The reserved seating, bowl seating and stadium seating are successfully classified by using RecGNN. RecGNN's weight parameter is optimized by the Lotus Effect Optimization Algorithm (LEA). On the Python working platform, the proposed RecGNN-LEA is implemented. The proposed strategy is analyzed using performance measures including recall, f1-score, accuracy, precision, and computation time. RecGNN's weight parameter is optimized by the Lotus Effect Optimization Algorithm (LEA). On the Python working platform, the proposed RecGNN-LEA is implemented. The proposed method is analyzed using performance measures including recall, f1-score, accuracy, precision, and computation time. The gained results of the proposed RecGNN-LEA method attains higher accuracy of 16.61%, 18.89%, and 17.92%, higher sensitivity of 16.37%, 12.23%, and 18.56% and higher precision of 14.81%, 16.79%, and 18.23%. The proposed SASCV-RecGNN-LEA method is compared with the existing methods such as SASCV-CNN, SASCV-LaRSA, and SASCV-TDNN models respectively.