Learning path generation involves the computation of learning trajectories to personalize academic instruction to prevent school problems. The Educational Planning Problem (EPP) considers generating personalized learning paths by scheduling activities that satisfy expected grades while minimizing plans makespan. In this work, we propose two scheduling models with two new activity duration functions incorporating student stress and sequence dependence to mimic learning effects and increase personalization. We also developed a metaheuristic algorithm to solve real-size instances composed of a greedy randomized adaptive search procedure enhanced by a variable neighborhood search. Experimental results validate the efficiency of our approach and the importance of considering stress in learning trajectories personalization.