Evaluation of Mean Absolute Error in Collaborative Filtering for Sparsity Users and Items on Female Daily Network

摘要

Female Daily Network is a beauty platform company that has a feature to share women's experiences of beauty items by rating them and talk about that. Female Daily Network doesn't have a recommendation engine for the next user will review the beauty item. Collaborative Filtering is a method to make a recommendation based on users' reviews of items. Sparsity users and sparsity items are problems to make a recommendation engine based on collaborative filtering. Comparing it with cut the sparsity users and items can be used to find the best recommendation engine. In this paper observe the cut of sparsity users and sparsity items to find the best Mean Absolute Error for the recommendation engine. Experimental research shows us that the best minimal times users give rating is 3 and the minimal item rating given is 3 also with Mean Absolute Error Value: 0.548. But MAE results not significant tell us if the cutting data affect the sparsity users and items. Because of the difference MAE, only 0.05 average from other results of the test and also fluctuating.