Recent advances in sensor technology have led to enhanced data acquisition capabilities in construction sites. A wealth of data are being collected from GPSequipped heavy vehicles for a wide range of monitoring, management, and analysis purposes. The availability of detailed GPS trajectory data has opened up new opportunities for modeling and simulation of real-world construction operations. One of the emerging areas in this regard is data-driven modeling and simulation, which is a modeling framework that attempts to automatically generate discrete-event simulation (DES) models based on a rich set of observed data as well as dynamically adapt the generated model to changes in data. Within the overall framework of automatically generating a DES simulation model for earthmoving operations, this paper focuses on developing methods to convert complex movement data collected from scrapers into a modeling element of activity-cycle diagram and activity scanning modeling paradigm-based DES system. Scraper changes travel routes at every cycle and its trip patterns (e.g., travel path and speed) are very difficult to generalize using a known parametric model (e.g., theoretical probability distribution), which in turn complicates the problem of automatic model generation. To deal with this issue, this paper proposes the use of relation between travel time and travel distance with regard to coefficient of variation measures, expressed in two separate distributions, to capture information needed to construct speed and path scenarios.