David Ben-Shimon

Recommender System from Personal Social Networks

Recommender systems are found in many modern web sites for applications such as recommending products to customers. In this paper we propose a new method for recommender system that employs the users’ social network in order to provide better recommendation for media items such as movies or TV shows. As part of this paper we develop a new paradigm for incorporating the feedback of the user’s friends. A field study that was conducted on real subjects indicates the strengths and the weaknesses of the proposed method compared to other simple and classic methods. The system is envisioned to function as a service for recommending personalized media (audio, video, print) on mobile phones, online media portals, sling boxes, etc. It is currently under development within Deutsche Telekom Laboratories - Innovations of Integrated Communication projects.

RecSys Challenge 2015 and the YOOCHOOSE Dataset

The 2015 ACM Recommender Systems Challenge offered the opportunity to work on a large-scale e-commerce dataset from a big retailer in Europe which is accepting recommender system as a service from YOOCHOOSE. Participants tackled the problem of predicting what items a user intends to purchase, if any, given a click sequence performed during an activity session on the e-commerce website. The challenge ran for seven months and was very successful, attracting 850 teams from 49 countries which submitted a total of 5,437 solutions. The winners of the challenge scored approximately 50% of the maximum score, which we considered as an impressive achievement. In this paper we provide a brief overview of the challenge and its results.

Establishing User Profiles in the MediaScout Recommender System

The MediaScout system is envisioned to function as personalized media (audio, video, print) service within mobile phones, online media portals, sling boxes, etc. The MediaScout recommender engine uses a novel stereotype-based recommendation engine. Upon the registration of new users the system must decide how to classify the new users to existing stereotypes. In this paper we present a method to achieve this classification through an anytime, interactive questionnaire, created automatically upon the generation of new stereotypes. A comparative study performed on the IMDB database illustrates the advantages of the new system

Anytime Algorithms for Recommendation Service Providers

Recommender systems (RS) can now be found in many commercial Web sites, often presenting customers with a short list of additional products that they might purchase. Many commercial sites do not typically have the ability and resources to develop their own system and may outsource the RS to a third party. This had led to the growth of a recommendation as a service industry, where companies, referred to as RS providers, provide recommendation services. These companies must carefully balance the cost of building recommendation models and the payment received from the e-business, as these payments are expected to be low. In such a setting, restricting the computational time required for model building is critical for the RS provider to be profitable. In this article, we propose anytime algorithms as an attractive method for balancing computational time and the recommendation model performance, thus tackling the RS provider problem. In an anytime setting, an algorithm can be stopped after any amount of computational time, always ensuring that a valid, although suboptimal, solution will be returned. Given sufficient time, however, the algorithm should converge to an optimal solution. In this setting, it is important to evaluate the quality of the returned solution over time, monitoring quality improvement. This is significantly different from traditional evaluation methods, which mostly estimate the performance of the algorithm only after its convergence is given sufficient time. We show that the popular item-item top-N recommendation approach can be brought into the anytime framework by smartly considering the order by which item pairs are being evaluated. We experimentally show that the time-accuracy trade-off can be significantly improved for this specific problem.

An ensemble method for top-N recommendations from the SVD

SVD suffers from computational limitation when delivering top-N items online.An ensemble algorithm for getting top-N items from the SVD results is proposed.The algorithm maps the items to the leaves of multiple compact trees offline.Users are assigned online to one leaf in each tree for obtaining their top-N items.The algorithm delivers faster and more accurate top-N items than the base SVD. Matrix factorization methods such as the singular value decomposition technique have become very popular in the area of recommender systems. Given a rating matrix as input, these techniques output two matrixes with lower dimensional space that represent the user and item features. The relevance of item i to user u is revealed by the score of the dot product between u vector of features and i vector of features. High scores indicate greater relevance. In order to deliver the best recommendations for a given user based on these latent features, one must obtain the list of scores of all the items for the given user and sort the resulting list. When the size of the catalogue is large, this phase consumes a large amount of computational time and cannot be done online. Another drawback with this approach is that once such a list is computed for a given user, it remains finite and it is impossible to incorporate within it new activities of the user. Hence, the use of such techniques is limited online.In this paper we propose an ensemble method for building a forest of trees offline, where each leaf in each tree is holding a unique set of item vectors. Once a user is engaged with the system, its vector is classified to one leaf in each one of the trees in the forest for conducting a dot product with the corresponding items. By using this method we compute online only a small number of dot products for a given user vector allowing us to quickly retrieve dynamic recommendations from the SVD, thereby presenting an alternative to the existing method which computes and caches all of the dot products among the items and users. The method maps the items to the leaves of multiple compact trees offline, each tree is a weak recommendation model, creating a forest of decision trees algorithm in which users that are assigned to these leaves online are likely to produce high dot product scores with the items that are already in the leaves. We demonstrate the effectiveness of the suggested ensemble method by applying it to three public datasets and comparing it to a state-of-the-art algorithm aimed at solving the problem.

Configuring and monitoring recommender system as a service

Many small and medium e-commerce retailers and publishers use recommender systems (RS) to personalize the website content. Many of them do not have an on premise solution for doing that, but rather contact a company that delivers the RS as a service to their website. The service is then responsible for collecting and storing the data, building recommendation models, and answering recommendation requests. Once the integration to such a service is done, the e-commerce retailer still wish to have some control on the service. Control that allows him to configure the recommendation models, turn off/on the service, apply filters on recommendations, define fallback models and more. In this demo we provide an overview of a real backend system which enables to a typical website owner exactly these capabilities. Capabilities for controlling the RS service in terms of configuration, management and monitoring.

Kernels for the Relevance Vector Machine - An Empirical Study

The Relevance Vector Machine (RVM) is a generalized linear model that can use kernel functions as basis functions. Experiments with the Matern kernel indicate that the kernel choice has a significant impact on the sparsity of the solution. Furthermore, not every kernel is suitable for the RVM. Our experiments indicate that the Matern kernel of order 3 is a good initial choice for many types of data.

In-House Solution for the RecSys Challenge 2015

RecSys Challenge 2015 is about predicting the items a user will buy in a given click session. We describe the in-house solution to the challenge as guided by the YOOCHOOSE team. The presented solution achieved 14th place in the challenges final leaderboard with a score of 51,932 points, while the winner obtained 63,102 points. We suggest two simple and easy to reconstruct approaches for obtaining a prediction in each session. In the first approach we suggest one classifier to determine whether each item in the session will be bought. In the second approach we suggest a two level classification model in which the first level determines whether the session is going to end with a purchase or not, and if it ends with a purchase, the second level classification determines the items that are going to be purchased.

Anytime algorithms for top-N recommenders

Many small and mid-sized e-businesses use the services of recommender system (RS) provider companies to outsource the construction and maintenance of their RS. The fees that RS providers charge their clients must cover the computation costs for constructing and updating the recommendation model. By using anytime algorithms, a RS provider can control the computation costs and still offer a system capable of delivering reasonable recommendations. Thus, a RS provider should be able to stop the construction of a recommendation model once the cost for compu-ting it reaches the amount the customer has agreed to pay. In this research we suggest anytime algorithms as a possible solu-tion to a problem that RS providers face. We demonstrate how certain existing recommendation algorithms can be adjusted to the anytime framework. We focus on the case of item-item algorithms, showing how the anytime behavior can be improved using different ordering methods of computations. We conduct a comparative study demonstrating the benefits of the proposed methods for top-N item-item recommenders.