Context. Today, the problem of visual simulation of passenger flow in public transport is essential in creating information systems for the development of modern Smart City. In Industry 4.0, it is crucial to develop technologies, means, and tools for implementing a single self-regulatory intelligent data exchange system in the provision of appropriate passenger transportation services in public transport. Today the following is essential: to visually display problem areas on routes in Smart City; to form and identify the main stops in time sections with the largest passenger exchange; to create proposals on the need to modernise routes taking into account the increase in public transport congestion in certain areas of Smart City, and to obtain results of passenger flow forecasting when making appropriate changes based on machine learning methods.
 Objective of the study is to develop a technology for visual simulation of passenger traffic in the field of public transport to improve the quality of passenger services in Smart City.
 Method. They have improved the simulation model for calculating passenger flow when changing the number of rolling stock on the route, in contrast to the known, added forecasting based on the developed neural network. The mechanism of visual simulation of passenger flows using GoogleMaps maps and dynamic movement on them with control of simulation display speed has been improved. A neural network with fully connected layers utilising an optimisation algorithm with an adaptive level of learning Adam to predict the flow of passengers between stops for a certain period of the day is proposed. Criteria for detailing passenger flows on urban routes are defined, including general indicators of the ratio of passenger traffic at a specific stop to the current period of the day. When designing the intelligent system, changing the capacity of public transport rolling stock in Smart City was further developed. Unlike the known ones, the available vehicles limit the change of power. The method of calculating a set of indicators of passenger traffic at stops and races, taking into account different local schedules and the specifics of transport on individual routes, has undergone further development.
 Results. An intelligent system of visual modelling of passenger traffic based on a neural network and machine learning has been developed, allowing optimising passenger traffic by public transport in Smart City. This data presentation makes it possible to assess the profitability of adding a new vehicle to the route or adjusting the schedule of other cars to cover the loaded areas during peak hours better. The well-known standard of public transport data presentation – GTFS is used for the operation of the software. It allows you to adapt the developed software product to the universal, rather than specific to a particular city or country. It was provided with a comparison of the obtained results on a data set of trolleybus routes (about 2000 records, collected based on experimental marketing research) in Lviv (Ukraine) to form a forecast of changes in passenger flow on certain sections at different times.
 Conclusions. It was found that the passenger flows predicted by the neural network in comparison with the actual ones lead to their growth by an average of 28% in critical races at rush hour. These results allow us to justify adding a schedule of a new vehicle for better coverage of loaded areas during peak hours. A comparison of changes in passenger traffic distributed by races during the day from 19:00 to 20:00, according to actual data and after the operation of the neural network indicates an increase in their average 70% of races that were predicted, which will allow a reasonable decision to launch additional transport on appropriate routes.