The technological evolutions have promoted mobile devices from rudimentary communication facilities to advanced personal assistants. According to the huge amount of accessible data, developing a time-saving and cost-effective method for location-based recommendations in mobile devices has been considered a challenging issue. This paper contributes a state-of-the-art solution for a personalized recommender assistant which suggests both accurate and unexpected point of interests (POIs) to users in each part of the day of the week based on their previously monitored, daily behavioral patterns. The presented approach consists of two steps of extracting the behavioral patterns from users’ trajectories and location-based recommendation based on the discovered patterns and user’s ratings. The behavioral pattern of the user includes their activity types in different parts of the day of the week, which is monitored via a combination of a stay point detection algorithm and an association rule mining (ARM) method. Having the behavioral patterns, the system exploits two recommendation procedures based on conventional collaborative filtering and K-furthest neighborhood model to recommend typical and serendipitous POIs to the users. The suggested POI list contains not only relevant and precise POIs but also unpredictable and surprising items to the users. To evaluate the system, the values of RMSE of each procedure were computed and compared. Conducted experiments proved the feasibility of the proposed solution.