Aggregated recommendation through random forests.

Area of Science:

• Computer Science
• Artificial Intelligence
• Machine Learning

Background:

• Traditional recommendation systems face challenges with new users or items (cold-start problem).
• Aggregated recommendation, suggesting items to user groups, offers a more general approach suitable for cold-start scenarios.

Purpose of the Study:

• To propose a novel random forest approach for building aggregated recommender systems.
• To predict group user ratings for items, addressing the cold-start recommendation challenge.

Main Methods:

• Constructed an aggregated decision table by merging user, item, and rating information.
• Modeled data conversion for new user, new item, and both new scenarios.
• Built a random forest where each leaf node contains a distribution of discrete ratings.
• Developed four prediction methods to calculate evaluation values from tree distributions.

Main Results:

• The proposed aggregated recommendation approach was evaluated on the MovieLens dataset.
• Experimental results indicate that the aggregated approach achieves acceptable accuracy levels.

Conclusions:

• The random forest-based aggregated recommendation system is a viable method for addressing cold-start recommendation problems.
• The approach demonstrates practical effectiveness in predicting group user ratings.