Author(s): Dr. T Srikanth, E Kavya, E Madhavi and K Shailaja

Abstract: The efficiency and quality of metro services may be improved with accurate predictions of Origin-Destination (OD) passenger flow. Projecting OD in metro networks has received less attention than projecting incoming as well as outgoing flows for particular stations, according to existing research. The problems with OD flows stem from three main sources: 1) their complicated geographical correlations and high temporal dynamics; 2) their susceptibility to external influences; and 3) the fact that their data slices are sparse and incomplete. In this study, we present an AFFN that can a) learn the effects of external factors automatically and b) precisely represent the periodic variations in passenger flows by fusing spatial dependencies from various based on knowledge graphs as well as hidden correlations between stations. Improving the accuracy of OD predictions is our secondary goal in extending AFFN to multi-task AFFN, which allows us to handle sparse and incomplete OD matrices by also predicting the input and output of each station. Two authentic Chinese metro trip datasets, gathered in Nanjing and Xi'an, were the subjects of our comprehensive experiments. When compared to state-of-the-art baseline approaches and AFFN variations, our AFFN and multitasking AFFN perform better on several accuracy criteria. This proves that AFFN and its components are useful in OD prediction.

DOI: 10.33545/26633582.2024.v6.i2b.129

Pages: 96-100 | Views: 1057 | Downloads: 652

Download Full Article: Click Here