I. Introduction
Experts and scholars have attempted to solve the increasingly prominent problem of urban traffic congestion via traffic flow prediction. With continuous exploration and development, traffic flow prediction has consequently achieved remarkable results in theoretical research and practical application. In traffic management and urban development planning, the laws and modes of vehicle flow should be clearly comprehended. Furthermore, traffic forecasting and management methods should be improved for managing urban traffic congestion. With the rapid development of big data and artificial intelligence technology since 2010, the research on traffic flow prediction has been conducted at varying degrees worldwide. In recent years, the European Commission has been discussing the various application problems of its member states’ relevant departments and governments. The European Commission also plans to financially support new projects and communicate with stakeholders through legislative means to ensure the smooth implementation of such endeavors [1]. As for the construction of systems, researchers generally use statistics, control theory, or data mining algorithms to achieve the core application of traffic network flow prediction and traffic state discrimination based on massive traffic data or parameters. Short-term traffic flow prediction and intelligent traffic guidance planning are the two urgent research directions of urban traffic network management [2]. Traffic flow prediction refers to the use of historical information, such as current traffic flow, weather conditions, environmental temperature and humidity, to predict future traffic flows. The essence of traffic flow prediction is the approximation and estimation of nonlinear traffic flows [3]. Traffic flow prediction can be divided into long-term and short-term predictions according to the size of time granularities. Long-term prediction refers to the prediction of the time series of traffic flow parameters in periods of large time granularity, whereas short-term traffic flow prediction entails much smaller granularities. The smaller is the time granularity, the more prominent are the instability characteristics of traffic flows. Therefore, short-term traffic flows are more difficult to predict than long-term traffic flows [4]. Given the large time steps of long-term prediction and numerous uncertain factors, a short time step between 5 and 15 minutes is often used for prediction [5]. At present, most studies regarding short-term traffic flow prediction use five minutes as the time granularity, whereas much fewer studies use one minute. Traffic flow prediction models include linear and nonlinear models [6]. A nonlinear prediction model refers to a model that uses neural network, deep learning, and other methods, including the short-term prediction model based on the Long Short-Term Memory (LSTM) neural network—a current research hotspot. This type of model relies on the powerful performance of currently existing computers; with the use of neural networks, short-term traffic flow prediction can achieve good accuracy [7], [8]. PTV-VISSIM is a micro-simulation modeling tool based on time interval and driving behavior, and it is used for the traffic modeling of urban traffic and public transport operation. Moreover, this tool is a discrete and random micro-simulation model with a time step of 100 s. The longitudinal motion of a vehicle adopts the psychophysical car-following model, whereas the lateral motion (lane change) adopts the rule-based algorithm. PTV-VISSIM can analyze the operation status of urban traffic and public transport under various traffic conditions, such as lane setting, traffic composition, traffic signal, bus stop, etc. Furthermore, as a tool, PTV-VISSIM can effectively evaluate traffic engineering designs and urban planning schemes.