A Chein-Shung hwang 0.4 0.3 0.2 HHGAFMCF AEWMCF 0.0 Figure 5. Recall-metric for various approaches vs recommendation number 0.4 0.3 ← GAWMCF 0.2 |— GAFMCF A-EWMCF 0.1 米SUCF 0.0 510152 0 Figure 6. Fl-metric for various approaches vs recommendation number 6. Conclusions fea In this paper, we have proposed a multi-criteria CF recommender system based on GAs for feature weighting. We treat the multi-criteria recommendations as optimization problems and apply a weighted average method by combining values from different criteria. The proposed approach first uses the traditional user-based CF algorithm to compute the prediction for each single criterion and then aggregates the overall prediction based on the weighting values derived by gas The GAs are implemented by two different approaches: filter and wrapper. Experimental results show that the wrapper-based approach is superior to all the other methods in all cases. The results also confirm that the provision of additional information about user preference in terms of different aspects for an item can potentially increase the recommendation quality. A further study for incorporating more information, such as contextual and content information, is under investigation 7. References 1 G. Adomavicius and A. Tuzhilin, Toward the next generation of recommender systems: A survey of the state-of-the-art and possible extensions", IEEE Transactions on Knowledge and Data Engineering, vol. 17 734-749.200 R. Burke, " Hybrid recommender systems: Survey and experiments, User Modeling and User Adapted Interaction, vol. 12, no. 4, pp 331-370, 2002 [3]M. Balabanovic and Y. Shoham,"Fab: Content based, collarative ree Communications of the ACM, vol. 40, no 3, pp 66-72, 1997Genetic Algorithms for Feature Weighting in Multi-criteria Recommender Systems Chein-Shung Hwang 0.0 0.1 0.2 0.3 0.4 3 5 10 15 20 25 30 Recall RN GAWMCF GAFMCF EWMCF SUCF Figure 5. Recall-metric for various approaches vs. recommendation number 0.0 0.1 0.2 0.3 0.4 3 5 10 15 20 25 30 F1 RN GAWMCF GAFMCF EWMCF SUCF Figure 6. F1-metric for various approaches vs. recommendation number 6. Conclusions In this paper, we have proposed a multi-criteria CF recommender system based on GAs for feature weighting. We treat the multi-criteria recommendations as optimization problems and apply a weighted average method by combining values from different criteria. The proposed approach first uses the traditional user-based CF algorithm to compute the prediction for each single criterion and then aggregates the overall prediction based on the weighting values derived by GAs. The GAs are implemented by two different approaches: filter and wrapper. Experimental results show that the wrapper-based approach is superior to all the other methods in all cases. The results also confirm that the provision of additional information about user preference in terms of different aspects for an item can potentially increase the recommendation quality. A further study for incorporating more information, such as contextual and content information, is under investigation. 7. References [1] G. Adomavicius and A. Tuzhilin, “Toward the next generation of recommender systems: A survey of the state-of-the-art and possible extensions”, IEEE Transactions on Knowledge and Data Engineering, vol. 17, no. 6, pp.734–749, 2005. [2] R. Burke, “Hybrid recommender systems: Survey and experiments”, User Modeling and User￾Adapted Interaction, vol. 12, no. 4, pp. 331–370, 2002. [3] M. Balabanovic and Y. Shoham, “Fab: Content based, collarative recommendation”, Communications of the ACM, vol. 40, no. 3, pp. 66-72, 1997. 134