2026-09-12T16:55:56.090Z http://ezaposleni.singidunum.ac.rs/rest/sciNaucniRezultati/oai

ezaposleni.singidunum.ac.rs/rest/sciNaucniRezultati/oai:2:12168 2026-09-07T23:39:20Z 2

Forecasting public transportation passenger load via metaheuristic optimized decomposition aided long short-term neural networks 2026 http://ezaposleni.singidunum.ac.rs/rest/sciNaucniRezultati/oai/record/2/12168 https://www.sciencedirect.com/science/article/pii/S2210539526002567 L. Jovanovic V. Simic D. Pamucar M. Zivkovic M. Antonijevic J. Perisic N. Bacanin Public transportation plays an important role in modern society. It helps decrease energy demands and pollution profiles of cities and helps provide better economic opportunities for communities. However, ensuring the proper function of public transportation requires significant experience and is not an easy task. Ensuring maximum utilization of transportation fleets while keeping idling costs low requires proper planning. A major challenge in this process is accurately forecasting passenger load on transportation routes and stations. This work presents a robust artificial intelligence-based approach aided by the variational mode decomposition techniques for forecasting the maximum monthly passenger load at individual public transportation stations. An altered version of an acknowledged metaheuristic is presented for the optimization of the decomposition process, alongside long short-term neural networks employed for forecasting. In order to manage decomposition optimization in the first layer and network optimization in the second, a two-layer architecture is presented. The introduced approach has been tested against several well-known metaheuristics when applied to a real-world public transportation dataset and has demonstrated promising outcomes. The top-performing models provide a significant level of stability, evaluated at the three largest public transportation hubs in Chicago City. The R2 scores of 0.857921 for Adams Wabash station, 0.939538 for Clark Lake station and 0.838688 for Washington Wells station were obtained. article 10.1016/j.rtbm.2026.101850 69 101850 2210-5409 Research in Transportation Business & Management