jinni.com Search base on your: Mood Time available Reviews Movie features Your movie history Semantic search
Which one you Choose and WHY?!
Current Best Approach: Collaborative Filtering with Temporal Dynamics: Time or history of users are Important because: People may changes their behavior or their preferences during their life. Capturing time drifting patterns in user behavior is essential to improving accuracy of recommenders. customer preferences change as new products and services become available
How Deep we should Consider in Time Time (history of user) Past Now
Current Best Approach: Collaborative Filtering with Temporal Dynamics: Overall Rating: ? 1 2 3 4 What is next!?
5 Star Rating!
5 Star Rating Global Rating: Local Rating: Some users just rate between 2-4 Some just 4,5 5 Star rating is the most unreliable rating feature!
Rating for What?! It’s Interesting! That movie is Excitement! I’m Like that movie! You may Interest in one movie, but after you watched it, you may don’t like that! or before you watched a movie you may know that you don’t like it, but because you have some interest in that movie you watched that.
Movie Features? Cast Actor Director Genre What else are missing?
Movie Class? Dynamic Features: Supports
Movies Supports? Supports Factors:
New Approach: Local Rating: 1 2 3 4 Class: A Class: B Class: A Class: C Find Similarities 1 2 3 4 1 2 3 4
New Approach: User: Each user is individually monitor and then base on their behavior/interests group them in clusters. Consider User Rating as Local and for overall rating convert local rating to global Movies: Each movie has a global ratingand Class(Static and Dynamic)