Feature selection for recommendation of movies

Recommender system is one of the recent applications which help to recommend the items or products based on the user needs. Since the user needs varies from person to person we cannot generalize the recommender system. Movie recommender system also has the same issues since individuals have different expectations while watching a movie and recommendation is not possible based on the annotations given by the other users. To overcome this situation an affective recommender framework is proposed in this work. Using the objectivity of audio-visual descriptor, connotation provides a space to predict the emotional state of the viewers. By extracting the connotative attributes such as audio-visual descriptor and user emotional state the connotative space is created. Then the movies which are nearer to each other in the created connotative space are recommended. Finally, the ability of the framework is assessed by employing the subjective analysis by asking the users to verify the film contents which met their affective requests.