The problem of modelling the lane-changing process in work zone areas is important. This study proposed a multistage analytic model based on driving simulator experiment. Several scenarios under different traffic volumes and lane-ends sign positions are tested to collect the behavioral variables including distance, speed, and acceleration. Three successive stages covering the initiation part (including one stage) and action part (including two stages) are defined to depict driver’s acceleration patterns with mathematical formulations. The model reduces the complexity of depicting the lane-changing behavior by using critical acceleration variables. Meanwhile, some important conclusions were found. High traffic volume condition is more prone to aggressive driving behavior which reflected in larger longitudinal and lateral acceleration. Additionally, drivers closer to work zone tend to take an abrupt turning since less merging time and larger lateral acceleration were accepted. These findings are useful in work zone planning and management.