Human behavior is an integral part of critical infrastructure dynamics, and therefore modeling the processes in which humans participate is crucial for the analytical capabilities of any digital twin of these infrastructures. Because of its variable and complex nature, however, using real data is imperative to the development of realistic simulation models of such processes. This work aims to develop a data-driven approach for pedestrian flow simulation modeling, by extracting and reconstructing processes in which the pedestrians take part. The proposed methods enable detection and reconstruction of processes from real-world data, and their integration into a pedestrian flow simulation model in an automated manner. In particular, the input data consists of spatio-temporal information about each pedestrian in the scenario, which could be obtained by location acquisition technologies such as GPS or camera tracking. For validation purposes, synthetic data was generated from a pedestrian movement scenario as initial input to the methods, with processes in the scenario modelled after those found in critical infrastructures of our application domain, such as a railway station or an airport. Our focus is to identify key processes that slow down or halt pedestrian movement, thereby detecting phenomena such as queueing, waiting before going to a counter, or going into a shop. The areas where these processes occur are referred to as points of interest [1], or stay point areas. First, we detect individual stay points for each agent [2] and then perform density-based clustering to obtain the areas. Furthermore, we extract properties of the processes within these areas, which can provide valuable information about how they should be modelled. These include individual process properties, such as stationarity of agent flow inside an area, or possible interactions between multiple processes, such as a causal relationship of the agent flow between areas- for example, an agent leaving a waiting area as soon as a counter becomes available. Based on their properties, the areas are then classified. Finally, we model the extracted processes based on the identified area types within a pedestrian flow simulation. We use a simulation framework based on the optimal steps model [3], which includes the functionality to model the processes, and therefore the complex dynamics of our target domain.