Diego Fernández

Comparison of collaborative filtering algorithms: Limitations of current techniques and proposals for scalable, high-performance recommender systems

The technique of collaborative filtering is especially successful in generating personalized recommendations. More than a decade of research has resulted in numerous algorithms, although no comparison of the different strategies has been made. In fact, a universally accepted way of evaluating a collaborative filtering algorithm does not exist yet. In this work, we compare different techniques found in the literature, and we study the characteristics of each one, highlighting their principal strengths and weaknesses. Several experiments have been performed, using the most popular metrics and algorithms. Moreover, two new metrics designed to measure the precision on good items have been proposed. The results have revealed the weaknesses of many algorithms in extracting information from user profiles especially under sparsity conditions. We have also confirmed the good results of SVD-based techniques already reported by other authors. As an alternative, we present a new approach based on the interpretation of the tendencies or differences between users and items. Despite its extraordinary simplicity, in our experiments, it obtained noticeably better results than more complex algorithms. In fact, in the cases analyzed, its results are at least equivalent to those of the best approaches studied. Under sparsity conditions, there is more than a 20% improvement in accuracy over the traditional user-based algorithms, while maintaining over 90% coverage. Moreover, it is much more efficient computationally than any other algorithm, making it especially adequate for large amounts of data.

Distributed architecture for k-nearest neighbors recommender systems

Collaborative filtering is one of the most popular recommendation techniques. While the quality of the recommendations has been significantly improved in the last years, most approaches present poor efficiency and scalability. In this paper, we study several factors that affect the performance of a k-Nearest Neighbors algorithm, and we propose a distributed architecture that significantly improves both throughput and response time. Two techniques for distributing recommender systems, user and item partition, were proposed and evaluated using that simulation model. We have found that user partition is generally better, with a faster response time and higher throughput.

Improving k-nearest neighbors algorithms: practical application of dataset analysis

In the last years, recommender systems have achieved a great popularity. Many different techniques have been developed and applied to this field. However, in many cases the algorithms do not obtain the expected results. In particular, when the applied model does not fit the real data the results are especially bad. This happens because many times models are directly applied to a domain without a previous analysis of the data. In this work we study the most popular datasets in the movie recommendation domain, in order to understand how the users behave in this particular context. We have found some remarkable facts that question the utility of the similarity measures traditionally used in k-Nearest Neighbors (kNN) algorithms. These findings can be useful in order to develop new algorithms. In particular, we modify traditional kNN algorithms by introducing a new similarity measure specially suited for sparse contexts, where users have rated very few items. Our experiments show slight improvements in prediction accuracy, which proves the importance of a thorough dataset analysis as a previous step to any algorithm development.

Advancing Network Flow Information Using Collaborative Filtering

Collaborative Filtering algorithms are frequently employed in e-commerce. However, this kind of algorithms can also be useful in other domains. In an information system thousands of bytes are sent through the network every second. Analyzing this data can require too much time and many resources, but it is necessary for ensuring the right operation of the network. Results are used for profiling, security analysis, traffic engineering and many other purposes. Nowadays, as a complement to a deep inspection of the data, it is more and more common to monitor packet flows, since it consumes less resources and it allows to react faster to any network situation. In a typical ow monitoring system, flows are exported to a collector, which stores the information before being analyzed. However, many collectors work based on time slots, so they do not analyze the flows when they are just received, generating a delay. In this work we demonstrate how Collaborative Filtering algorithms can be applied to this new domain. ...

Using Collaborative Filtering in a new domain: traffic analysis

The importance of information systems is increasing every day. In order to ensure their right operation, it is necessary to analyze a huge amount of traffic generated by different devices. However, classical techniques for operation and management are reactive and not proactive, what can evolve in a failure in the system. In this work we propose a new approach where we analyze network traffic using Collaborative Filtering. In other domains, these systems have proved to filter thousands of items according to user needs and tastes. They can predict user preferences and recommend relevant items for the user. In this sense, in this new domain, relevant items are data flows, so our goal is to recommend flows which are related to the traffic already captured.