Performance Prediction for Quality Recommendations

Work in the area of collaborative filtering continues to predominantly focus on prediction accuracy as a measure of the quality of the systems. Other measures of quality of these systems have been explored but not to the same extent. The work described in this chapter considers quality from the perspective of performance prediction. Per user, the performance of a collaborative filtering system is predicted based on rules learned by a machine learning approach. The experiments outlined aim, using three different datasets, to firstly learn the rules for performance prediction and to secondly test the accuracy of the rules produced. Results show good performance prediction accuracy can be found for all three datasets. The work does not step too far from the idea of prediction accuracy as a measure of quality but it does consider prediction accuracy from a different perspective, that of predicting the performance of a collaborative filtering system, per user, in advance of recommendation.