This article investigates the impact of rating data characteristics on the performance of several popular recommendation algorithms, including user-based and item-based collaborative filtering, as well as matrix factorization. We focus on three groups of data characteristics: rating space, rating frequency distribution, and rating value distribution. A sampling procedure was employed to obtain different rating data subsamples with varying characteristics; recommendation algorithms were used to estimate the predictive accuracy for each sample; and linear regression-based models were used to uncover the relationships between data characteristics and recommendation accuracy. Experimental results on multiple rating datasets show the consistent and significant effects of several data characteristics on recommendation accuracy.

@article{adomavicius2012impact, author = {Adomavicius, Gediminas and Zhang, Jingjing}, title = {Impact of Data Characteristics on Recommender Systems Performance}, journal = {ACM Trans. Manage. Inf. Syst.}, publisher = {ACM}, address = {New York, NY, USA}, year = {2012}, volume = {3}, number = {1}, pages = {3:1--3:17}, url = {http://doi.acm.org/10.1145/2151163.2151166}, doi = {10.1145/2151163.2151166}, issn = {2158-656X}, keywords = {characteristics, dataset, dependence, evaluation, model, recommender}, abstract = {This article investigates the impact of rating data characteristics on the performance of several popular recommendation algorithms, including user-based and item-based collaborative filtering, as well as matrix factorization. We focus on three groups of data characteristics: rating space, rating frequency distribution, and rating value distribution. A sampling procedure was employed to obtain different rating data subsamples with varying characteristics; recommendation algorithms were used to estimate the predictive accuracy for each sample; and linear regression-based models were used to uncover the relationships between data characteristics and recommendation accuracy. Experimental results on multiple rating datasets show the consistent and significant effects of several data characteristics on recommendation accuracy.} }

%0 = article %A = Adomavicius, Gediminas and Zhang, Jingjing %C = New York, NY, USA %D = 2012 %I = ACM %T = Impact of Data Characteristics on Recommender Systems Performance %U = http://doi.acm.org/10.1145/2151163.2151166

We present a new class of statistical de- anonymization attacks against high-dimensional micro-data, such as individual preferences, recommendations, transaction records and so on. Our techniques are robust to perturbation in the data and tolerate some mistakes in the adversary's background knowledge. We apply our de-anonymization methodology to the Netflix Prize dataset, which contains anonymous movie ratings of 500,000 subscribers of Netflix, the world's largest online movie rental service. We demonstrate that an adversary who knows only a little bit about an individual subscriber can easily identify this subscriber's record in the dataset. Using the Internet Movie Database as the source of background knowledge, we successfully identified the Netflix records of known users, uncovering their apparent political preferences and other potentially sensitive information.

@inproceedings{narayanan2008robust, author = {Narayanan, Arvind and Shmatikov, Vitaly}, title = {Robust De-anonymization of Large Sparse Datasets}, booktitle = {Proc. of the 29th IEEE Symposium on Security and Privacy}, publisher = {IEEE Computer Society}, year = {2008}, pages = {111--125}, url = {http://www.cs.utexas.edu/~shmat/shmat_oak08netflix.pdf}, doi = {10.1109/SP.2008.33}, keywords = {anonymization, datamining, dataset, netflix, privacy, recommender, toread}, abstract = {We present a new class of statistical de- anonymization attacks against high-dimensional micro-data, such as individual preferences, recommendations, transaction records and so on. Our techniques are robust to perturbation in the data and tolerate some mistakes in the adversary's background knowledge. We apply our de-anonymization methodology to the Netflix Prize dataset, which contains anonymous movie ratings of 500,000 subscribers of Netflix, the world's largest online movie rental service. We demonstrate that an adversary who knows only a little bit about an individual subscriber can easily identify this subscriber's record in the dataset. Using the Internet Movie Database as the source of background knowledge, we successfully identified the Netflix records of known users, uncovering their apparent political preferences and other potentially sensitive information.} }

%0 = inproceedings %A = Narayanan, Arvind and Shmatikov, Vitaly %B = Proc. of the 29th IEEE Symposium on Security and Privacy %D = 2008 %I = IEEE Computer Society %T = Robust De-anonymization of Large Sparse Datasets %U = http://www.cs.utexas.edu/~shmat/shmat_oak08netflix.pdf

@inproceedings{1458098, author = {Song, Yang and Zhang, Lu and Giles, C. Lee}, title = {A sparse gaussian processes classification framework for fast tag suggestions}, booktitle = {CIKM '08: Proceeding of the 17th ACM conference on Information and knowledge mining}, publisher = {ACM}, address = {New York, NY, USA}, year = {2008}, pages = {93--102}, url = {http://portal.acm.org/citation.cfm?id=1458098}, doi = {http://doi.acm.org/10.1145/1458082.1458098}, isbn = {978-1-59593-991-3}, keywords = {bibsonomy, bookmarking, classification, dataset, ml, recommender, social, tag, tagging, taggingsurvey, toread} }

%0 = inproceedings %A = Song, Yang and Zhang, Lu and Giles, C. Lee %B = CIKM '08: Proceeding of the 17th ACM conference on Information and knowledge mining %C = New York, NY, USA %D = 2008 %I = ACM %T = A sparse gaussian processes classification framework for fast tag suggestions %U = http://portal.acm.org/citation.cfm?id=1458098

We present a new class of statistical de-anonymization attacks against high-dimensional micro-data, such as individual preferences, recommendations, transaction records and so on. Our techniques are robust to perturbation in the data and tolerate some mistakes in the adversary's background knowledge.

@misc{narayanan-2006, author = {Narayanan, Arvind and Shmatikov, Vitaly}, title = {How To Break Anonymity of the Netflix Prize Dataset}, year = {2006}, url = {http://www.citebase.org/abstract?id=oai:arXiv.org:cs/0610105}, keywords = {Preis, anonymity, dataset, netflix, prize, recommender}, abstract = { We present a new class of statistical de-anonymization attacks against high-dimensional micro-data, such as individual preferences, recommendations, transaction records and so on. Our techniques are robust to perturbation in the data and tolerate some mistakes in the adversary's background knowledge.} }

%0 = misc %A = Narayanan, Arvind and Shmatikov, Vitaly %B = } %C = %D = 2006 %I = %T = How To Break Anonymity of the Netflix Prize Dataset} %U = http://www.citebase.org/abstract?id=oai:arXiv.org:cs/0610105