Collaborative filtering is valuable in e-commerce, and for direct recommendations for music, movies, news etc. But today’s systems use centralized databases and have several disadvantages, including privacy risks. As we move toward ubiquitous computing, there is a great potential for individuals to share all kinds of information about places and things to do, see and buy, but the privacy risks are severe. In this paper we introduce a peer-to-peer protocol for collaborative filtering which protects the privacy of individual data. A second contribution of this paper is a new collaborative filtering algorithm based on factor analysis which appears to be the most accurate method for CF to date. The new algorithm has other advantages in speed and storage over previous algorithms. It is based on a careful probabilistic model of user choice, and on a probabilistically sound approach to dealing with missing data. Our experiments on several test datasets show that the algorithm is more accurate than previously reported methods, and the improvements increase with the sparseness of the dataset. Finally, factor analysis with privacy is applicable to other kinds of statistical analyses of survey or questionaire data scientists (e.g. web surveys or questionaires).

Server-based collaborative filtering systems have been very successful in e-commerce and in direct recommendation applications. In future, they have many potential applications in ubiquitous computing settings. But today's schemes have problems such as loss of privacy, favoring retail monopolies, and with hampering diffusion of innovations. We propose an alternative model in which users control all of their log data. We describe an algorithm whereby a community of users can compute a public "aggregate" of their data that does not expose individual users' data. The aggregate allows personalized recommendations to be computed by members of the community, or by outsiders. The numerical algorithm is fast, robust and accurate. Our method reduces the collaborative filtering task to an iterative calculation of the aggregate requiring only addition of vectors of user data. Then we use homomorphic encryption to allow sums of encrypted vectors to be computed and decrypted without exposing individual data. We give verification schemes for all parties in the computation. Our system can be implemented with untrusted servers, or with additional infrastructure, as a fully peer-to-peer (P2P) system. 1

In this paper we report on a pilot user study aimed at evaluating two aspects of recommender systems that have not been the aim of previous user studies in the field. Firstly, item semantics may be incorporated into a collaborative recommender system and we wish to measure the effect on user satisfaction. Secondly, we would like to evaluate different approaches to collecting ratings from users: the ratings that are used to seed their profile with a collaborative filtering system. Key indications from the study are: users do prefer recommendations generated by semantically enhanced recommender systems; the user’s satisfaction with a recommendation set is different from the sum of their satisfaction with the individual items with the set and the approach to collecting item ratings from the user should be tailored to the algorithm being used. Finally, recommender systems within the movie domain seem to be more useful for “movie buffs ” rather than the “average movie watcher ” for whom recommending simply the most popular movies seems to be most appropriate. 1

Abstract Collaborative Filtering (CF) is a popular method for personalizing product recommendations for e-Commerce and customer relationship management (CRM). CF utilizes the explicit or implicit product evaluation ratings of customers to develop personalized recommendations. However, there has been no in-depth investigation of the parameters of CF in relation to the number of ratings on the part of an individual customer and the total number of ratings for an item. We empirically investigated the relationships between these two parameters and CF performance, using two publicly available data sets, EachMovie and MovieLens. We conducted three experiments. The first two investigated the relationship between a particular customer’s number of ratings and CF recommendation performance. The third experiment evaluated the relationship between the total number of ratings for a particular item and CF recommendation performance. We found that there are ratings thresholds below which recommendation performance increases monotonically, i.e., when the numbers of customer and item ratings are below threshold levels, CF recommendation performance is affected. In addition, once rating numbers surpass threshold levels, the value of each rating decreases. These results may facilitate operational decisions when applying CF in practice.