You are hereAdaptive Resonance Theory-based Recommendation System
Adaptive Resonance Theory-based Recommendation System
This is my graduate project for my Artificial Intelligence class at Digipen. It uses Adaptive Resonance Theory (ART) to create a recommendation system. Specifically, in this case, a survey was created with a list of 20 chocolates, and respondents were asked to rate each chocolate from 0-10, 0 meaning never tried, 1 meaning tried but disliked, and 10 being tried and liked a lot. The recommendation system takes the survey data, clusters users with similar chocolate tastes together, and tries to make a recommendation to users for chocolates that other people with similar tastes have liked.
The implementation itself is based on M Tim Jones' chapter on Adaptive Resonance Theory in AI Application Programming from Charles River Media. It does however have 2 specific enhancements: the ability to cope with non-negative integral vectors as input (as opposed to just binary vectors) and the ability to group people into hyper-clusters. Although this system in particular is meant to recommend chocolates to respondents of a survey, the techniques and code involved are generic enough that it should be applicable towards other uses.
Click here to download the source for Adaptive Resonance Theory-based Recommendation System. The code is written in C#