• recommender system;
  • collaborative filtering;
  • case-based reasoning;
  • association;
  • emotion


Recommender systems have been widely accepted across a broad range of application areas. Their key task is to estimate the user's need as accurately as possible. To do so, conventional recommender systems have focused on how to more accurately determine similar users or similar products. However, individuals often have unique preferences, or they do not always prefer to use the same service. Typically, an individual's association from his own experience will be preferable, that is more accurate, compared with common interests he may have with other users. However, recommendation methods based on association, versus common interests, have been very few. Hence, the purpose of this paper is to propose an association model based reasoning methodology that estimates users’ associative strength from self-experiences based on association theories. In particular, we adopt the Rescorla–Wagner model, the Galton free association test, and Russell's circumflex model for mood positivity to estimate associative strength. We perform an experimental study with actual experience data to verify the hypotheses, which validates the association model based algorithm in finding the best service with a good elapsed time.