A Novel Recommendation Model Regularized with User Trust and Item Ratings
A Novel Recommendation Model Regularized with User Trust and Item Ratings
Abstract
We propose Trust SVD, a trust-based matrix factorization technique for recommendations. Trust SVD integrates multiple information sources into the recommendation model in order to reduce the data sparsity and cold star t problems and their degradation of recommendation performance. An analysis of social trust data from four real-world data sets suggests that not only the explicit but also the implicit influence of both ratings and trust should be taken into consideration in a recommendation model. Trust SVD therefore builds on top of a state-of-the-art recommendation algorithm, SVD++ (which uses the explicit and implicit influence of rated items), by fur there incorporating both the explicit and implicit influence of trusted and trusting users on the prediction of items for an active user. The proposed technique is the first to extend SVD++ with social trust information. Experimental results on the four data sets demonstrate
that Trust SVD achieves better accuracy than other ten counter parts recommendation techniques.
Existing System
Collaborative filtering (CF) is one of the most popular techniques to implement a recommender system. The idea of CF is that users with similar preferences in the past are likely to favor the same items (e.g., movies, music, books, etc.) in the future. CF has also been applied to tasks besides item recommendations, in domains such as image processing and bioinformatics. However, CF suffers from two well known issues: data sparsity and cold start. The former issue refers to the fact that users usually rate only a small portion of items, while the latter indicates that new users only give a few ratings (a.k.a. cold-start users). Both issues severely degrade the efficiency of a recommender system in modeling user preferences and thus the accuracy of predicting a user’s rating for an unknown item. To help resolve these issues, many researchers attempt to incorporate social trust information into their recommendation models, given that model-based CF approaches outperform memory-based approaches. These approaches further regularize the user-specific feature vectors by the phenomenon that friends often influence each other in recommending items. However, even the best performance reported by the latest work can be inferior to that of other state-of-the-art models which are merely based on user–item ratings. For instance, a well-performing trust-based model obtains 1.0585 on data set Epinions. com in terms of Root Mean Square Error (RMSE), whereas the performance of a user–item baseline (see, Koren, Section 2.1) can achieve 1.0472 in terms of RMSE.
Disadvantages:
- CF suffers from two well known issues are data sparsity and cold start.
Proposed System
We propose a novel trust-based recommendation model regularized with user trust and item ratings, termed Trust SVD. Our approach builds on top of a state-of the-art model SVD++ through which both the explicit and implicit influence of user–item ratings are involved to generate predictions. In addition, we further consider the influence of user trust (including trustees and trusters) on the rating prediction for an active user. To the authors’ knowledge, our work is the first to extend SVD++ with social trust information. Specifically, on one hand the implicit influence of trust (who trusts whom) can be naturally added to the SVD++ model by extending the user modeling. On the other hand, the explicit influence of trust (trust values) is used to constrain that user-specific vectors should conform to their social trust relationships. This ensures that user-specific vectors can be learned from their trust information even if a few or no ratings are given. In this way, the concerned issues can be better alleviated. Our method is novel for its consideration of both the explicit and implicit influence of item ratings and of user trust. In addition, a weighted-regularization technique is used to help avoid over-fitting for model learning.
Advantages
- In high-performing ratings-only models in terms of predictive accuracy, and is more capable of coping with the cold-start situations.
- To propose a novel trust based recommendation approach (TrustSVD2) that incorporates both (explicit and implicit) influence of rating and trust information.
System Architecture
Fig. The influence of (a) trustees v and (b) trusters k on the rating prediction for the active user u and target item j.
Modules
- Linear combination
- All as trusting users
- All as trusted users
Module Description
- Linear combination
A natural and straightforward way is to linearly combine the two kinds of implicit trust influence Specifically, means that we only consider the influence of trusting users; indicates that only the influence of trusted users are considered; and mixes the two kinds of trust influence together.
- All as trusting users
In a trust relationship, a user u can be represented either by trustor or trustee. An alternative way is to model the influence of users trust neighbors, including both trusted and trusting users, in the manner of trusting users.
- All as trusted users
With the same assumption, we may model the influence of all trust neighbors in the manner of trusted users However, since user-feature matrix P plays a key role in bridging both rating and trust information, the rating prediction.
Configuration:-
H/W System Configuration:-
Processor – Pentium –III
Speed – 1.1 Ghz
RAM – 256 MB(min)
Hard Disk – 20 GB
Floppy Drive – 1.44 MB
Key Board – Standard Windows Keyboard
Mouse – Two or Three Button Mouse
Monitor – SVGA
S/W System Configuration:-
Operating System : Windows95/98/2000/XP
Programming Language : Java
Comments are closed.