A novel recommender system for PROFIT
The PROFIT platform incorporates a novel hybrid recommender system that is going to provide recommendations to each user according to their specific interests, e.g. financial news and forecast reports that would appeal to them, users with similar interests, etc.
Personalized recommendations are based on user preferences as they are expressed through their explicit ratings of items in the PROFIT platform, as well as through their interactions with the platform that implicitly indicate their interests, such as posting articles on particular topics, commenting on specific articles, participating in polls, etc.
Let’s see a working example of the hybrid recommender system, based on data currently available on the platform. Consider user Thomas who has just registered to the PROFIT platform. After logging in, he checks the Recommendations tab, available on his user dashboard. Given that he is new to the system, these recommendations are based on the most popular items in the platform.
Next, Thomas glances quickly over the recommended articles and finds an article that attracts his attention. He indeed finds it interesting and gives it a 5-star rating. As information about his preferences is now available to the platform, personalized recommendations can be provided to him, using a hybrid scheme that combines content-based and collaborative filtering approaches. It should be noted that the collaborative filtering recommendations are based on the ratings of similar users in the platform and currently there are other users that have also rated the aforementioned article.
The hybrid recommender system manages to successfully rank in the top positions recommendations from both approaches, including items that would not have been provided using any of the individual approaches alone. However, more emphasis has been placed on the user-based collaborative filtering approaches given the fact that they are considered as more beneficial for systems with fewer users than items, such as PROFIT. User evaluations will determine what recommendations users find more appealing and will allow us to improve our approach!
You can read more about PROFIT’s recommender system here.