Adaptive nonnegative matrix factorization and measure comparisons for recommender systems

Abstract

The Nonnegative Matrix Factorization (NMF) of the rating matrix has shown to be an effective method to tackle the recommendation problem. In this paper we propose new methods based on the NMF of the rating matrix and we compare them with some classical algorithms such as the SVD and the regularized and unregularized non-negative matrix factorization approach. In particular a new algorithm is obtained changing adaptively the function to be minimized at each step, realizing a sort of dynamic prior strategy. Another algorithm is obtained modifying the function to be minimized in the NMF formulation by enforcing the reconstruction of the unknown ratings toward a prior term. We then combine different methods obtaining two mixed strategies which turn out to be very effective in the reconstruction of missing observations. We perform a thoughtful comparison of different methods on the basis of several evaluation measures. We consider in particular rating, classification and ranking measures showing that the algorithm obtaining the best score for a given measure is in general the best also when different measures are considered, lowering the interest in designing specific evaluation measures. The algorithms have been tested on different datasets, in particular the 1M, and 10M MovieLens datasets containing ratings on movies, the Jester dataset with ranting on jokes and Amazon Fine Foods dataset with ratings on foods. The comparison of the different algorithms, shows the good performance of methods employing both an explicit and an implicit regularization scheme. Moreover we can get a boost by mixed strategies combining a fast method with a more accurate one.

Maximizing E-Tailers’ Sales Volume through the Shipping-Fee Discount and Product Recommendation System

“…Different from the CF and CBF methods, which are based on the possibility of buying the product, other research focuses on the impact of the recommendation system on the customer's purchase decision. e researchers also explored the reaction of customers to the recommended products [35] and identified the most appropriate products to recommend [36,37]. In addition to customers' preferences, customers' savings, and the e-tailer's profits have emerged.…”

Prediction of Opinion Keywords and Their Sentiment Strength Score Using Latent Space Learning Methods

“…Incorporating the latent factors associated with users has been proven to be very useful to design effective algorithms in ranking and recommendation systems [8,28,29]. These works are based on the idea of factorizing a matrix to linearly reduce the dimensionality of the problem and extract some commonalities that may be implicit in the data under analysis.…”

“…Moreover, the large sparsity of the data under analysis can be up to 99% for datasets such as Amazon product review undermining the capabilities of this type of learning methods based on matrix decomposition. Some alternative works propose the use of specific regularization parameters to minimize this effect [28].…”

AANMF: Attribute-Aware Attentional Neural Matrix Factorization

“…Some works were proposed to enhance the performance of the MF model. Del Corso et al [6] built an NMF model that change adaptively the function to be minimized at each step. Wang et al [23] proposed an MF model named LOD-MF, which dug out implicit feedback information and applied a hybrid similarity measure to identify the semantically similar neighbors of the target item.…”