Similar to the item-based and user-based recommender systems explained earlier, we can also use model-based recommender implementations in Mahout, such as
SVDRecommender, which uses matrix factorization methods to generate recommendations.
The steps are similar to previous implementations. Two important steps that need to be understood here are as follows:
org.apache.mahout.cf.taste.impl.recommender.svd.ALSWRFactorizerclass, which factorizes the user rating matrix using Alternating-Least-Squares with Weighted-λ-Regularization. The
ALSWRFactorizerclass constructor takes parameters such as DataModel, the number of features, the regularization parameter, and the number of iterations as inputs. This
ALSWRFactorizerclass instance ...