Location-based social networks (LBSNs) become an essential part of our lives as these services can assist users in finding interesting point-of-interest (POI). Many studies have been conducted to perform POI recommendations with various factors, such as user's check-in records, geographic information, and social relationship. However, existing studies use only fixed values without considering the user's current situations that frequently change. In this paper, we propose a flexible POI recommendation based on the user's situation. For this, we first construct a graph model of three factors, such as trajectory, distance, and preference. We then use a weight-threshold algorithm (w-TA) to adjust the weight of each factor based on the user's current situation. The experiment results show the flexibility of the proposed method in various conditions.