Experimental Evaluation on Matrix Decomposition Methods
- 1.4k Downloads
In this chapter, we study the performance of described SVD and UV decomposition algorithms, against an improved version of the original item based CF algorithm combined with SVD. For the UV decomposition method, we will present the appropriate tuning of parameters of its objective function to have an idea of how we can get optimized values of its parameters. We will also answer the question if these values are generally accepted or they should be different for each data set. The metrics we will use are root-mean-square error (RMSE), precision, and recall. The size of a training set is fixed at 75 %, and we perform a fourfold cross-validation.
KeywordsExperiments SVD decomposition UV decomposition
Unable to display preview. Download preview PDF.
- 5.McLauglin, R., Herlocher, J.: A collaborative filtering algorithm and evaluation metric that accurately model the user experience. In: Proceedings of ACM SIGIR Conference, pp. 329–336 (2004)Google Scholar