As part of the Digital Single Market Strategy, the EU Commission focuses mainly on Intelligent Transportation Systems (ITS) solutions, for more efficient traffic management strategies (European Commission: Mobility and Transport, 2023). Within the next few years, traffic flow predictions will be used for most of the smart solutions in transportation engineering. Traffic Flow Prediction Models (TFPM) are recognized as a part of the foundations of ITS and many other real-time applications to provide quick and cost-effective solutions for reducing congestion and travel time. TFPM are the models that can estimate traffic flow at a particular location over a given period. These estimations can be carried out based on various techniques (like statistical and deep learning) which have been researched for decades to achieve more accuracy and faster computation. To achieve more flexibility and accuracy, more focus was given towards the deep learning models in the last few years. Even though the accuracy was achieved with such neural network algorithms, the complexity of the model and the processing time were also increasing. Two recent time series prediction models are Fb-Prophet and NeuralProphet. Especially NeuralProphet model which is the successor of the Fb-Prophet model was developed to bridge the traditional and deep learning models.
These models have already started dominating in various other fields for solving current prediction problems. Therefore, being new to traffic flow predictions, this thesis focuses on analysing the model’s performances by comparing with a well-known statistical prediction model namely - Seasonal Auto-regressive Integrated Moving Average model (SARIMA) and their applicability for an urban traffic scenario is studied. The property of decomposability of Fb-Prophet and NeuralProphet model enhances them with more flexibility to include the effect of holidays and events. The analyst in the loop concept makes the process of prediction faster and more automated. Both the models can be scaled up to a bigger network because of its automated functioning. 
Initially, the performances of these time series models are evaluated in terms of accuracy and computational time. Additionally, the prediction of urban traffic with inclusion of effect of external factors like weather and holidays are carried out by using NeuralProphet model. Thus, the predicted results will be given to a flow generator that can give SUMO readable traffic demand file. Finally, this thesis also outlines the development of a real-time simulation system that can be used for real-time applications, driving simulators, atypical scenario simulations and various other applications. A sample simulation for platoon formation was also run with the developed simulation system. The developed prediction system can be in future used for many research purposes. The predicted results from the model with effect of rainfall can also be simulated in a dynamic driving simulator and used for analysis of vehicle dynamics at various road surface conditions (wet/icy).