On Affinity Measures for Artificial Immune System Movie Recommenders

TL;DR

This paper investigates the impact of different affinity measures on the performance of Artificial Immune System-based movie recommenders, finding Weighted Kappa to be more effective than Kendall's Tau.

Contribution

It introduces a comparison of affinity measures within AIS for movie recommendation, highlighting Weighted Kappa's suitability and robustness of AIS with proper affinity selection.

Findings

Weighted Kappa outperforms Kendall's Tau for affinity measurement

AIS recommenders are robust with appropriate affinity measures

Proper affinity choice leads to good recommendation results

Abstract

We combine Artificial Immune Systems 'AIS', technology with Collaborative Filtering 'CF' and use it to build a movie recommendation system. We already know that Artificial Immune Systems work well as movie recommenders from previous work by Cayzer and Aickelin 3, 4, 5. Here our aim is to investigate the effect of different affinity measure algorithms for the AIS. Two different affinity measures, Kendalls Tau and Weighted Kappa, are used to calculate the correlation coefficients for the movie recommender. We compare the results with those published previously and show that Weighted Kappa is more suitable than others for movie problems. We also show that AIS are generally robust movie recommenders and that, as long as a suitable affinity measure is chosen, results are good.