We have developed a method for recommending items that combines content and collaborative data under a single probabilistic framework. We benchmark our algorithm against a naive Bayes classifier on the cold-start problem, where we wish to recommend items that no one in the community has yet rated. We systematically explore three testing methodologies using a publicly available data set, and explain how these methods apply to specific real-world applications. We advocate heuristic recommenders when benchmarking to give competent baseline performance. We introduce a new performance metric, the CROC curve, and demonstrate empirically that the various components of our testing strategy combine to obtain deeper understanding of the performance characteristics of recommender systems. Though the emphasis of our testing is on cold-start recommending, our methods for recommending and evaluation are general. Methods and Metrics for Cold-Start Recommendations ABSTRACTWe have developed a method for recommending items that combines content and collaborative data under a single probabilistic framework. We benchmark our algorithm against a naïve Bayes classifier on the cold-start problem, where we wish to recommend items that no one in the community has yet rated. We systematically explore three testing methodologies using a publicly available data set, and explain how these methods apply to specific real-world applications. We advocate heuristic recommenders when benchmarking to give competent baseline performance. We introduce a new performance metric, the CROC curve, and demonstrate empirically that the various components of our testing strategy combine to obtain deeper understanding of the performance characteristics of recommender systems. Though the emphasis of our testing is on cold-start recommending, our methods for recommending and evaluation are general.

The Internet is an important source of health information. Thus, the frequency of Internet searches may provide information regarding infectious disease activity. As an example, we examined the relationship between searches for influenza and actual influenza occurrence. Using search queries from the Yahoo! search engine ( http://search.yahoo.com ) from March 2004 through May 2008, we counted daily unique queries originating in the United States that contained influenza-related search terms. Counts were divided by the total number of searches, and the resulting daily fraction of searches was averaged over the week. We estimated linear models, using searches with 1-10-week lead times as explanatory variables to predict the percentage of cultures positive for influenza and deaths attributable to pneumonia and influenza in the United States. With use of the frequency of searches, our models predicted an increase in cultures positive for influenza 1-3 weeks in advance of when they occurred (P < .001), and similar models predicted an increase in mortality attributable to pneumonia and influenza up to 5 weeks in advance (P < .001). Search-term surveillance may provide an additional tool for disease surveillance.

Recent work has demonstrated that Web search volume can "predict the present," meaning that it can be used to accurately track outcomes such as unemployment levels, auto and home sales, and disease prevalence in near real time. Here we show that what consumers are searching for online can also predict their collective future behavior days or even weeks in advance. Specifically we use search query volume to forecast the opening weekend box-office revenue for feature films, first-month sales of video games, and the rank of songs on the Billboard Hot 100 chart, finding in all cases that search counts are highly predictive of future outcomes. We also find that search counts generally boost the performance of baseline models fit on other publicly available data, where the boost varies from modest to dramatic, depending on the application in question. Finally, we reexamine previous work on tracking flu trends and show that, perhaps surprisingly, the utility of search data relative to a simple autoregressive model is modest. We conclude that in the absence of other data sources, or where small improvements in predictive performance are material, search queries provide a useful guide to the near future.culture | predictions A s people increasingly turn to the Internet for news, information, and research purposes, it is tempting to view online activity at any moment in time as a snapshot of the collective consciousness, reflecting the instantaneous interests, concerns, and intentions of the global population (1, 2). From this perspective, it is a short step to conclude that what people are searching for today is predictive of what they will do in the near future. Consumers contemplating buying a new camera may search to compare models; moviegoers may search to determine the opening date of a new film, or to locate cinemas showing it; and individuals planning a vacation may search for places of interest, to find airline tickets, or to price hotel rooms. If so, it follows that by appropriately aggregating counts of search queries related to retail activity, moviegoing, or travel, one might be able to predict collective behavior of economic, cultural, or political interest. Determining the nature of behavior that can be predicted using search, the accuracy of such predictions, and the time scale over which predictions can be usefully made are therefore all questions of interest.Although previous work has considered the relation between search volume and offline outcomes, researchers have focused on the observation that search "predicts the present" (3, 4), meaning that search volume correlates with contemporaneous events. (8) showed that search volume for handpicked influenza-related queries was correlated with subsequently reported caseloads over the period 2004-2008, and Hulth et al. (9) found similar results in a study of search queries submitted on a Swedish medical Web site. An automated procedure for identifying informative queries is described in Ginsberg et al. (10), and based on that methodology, Google Flu Trend...

As a whole, the World Wide Web displays a striking ''rich get richer'' behavior, with a relatively small number of sites receiving a disproportionately large share of hyperlink references and traffic. However, hidden in this skewed global distribution, we discover a qualitatively different and considerably less biased link distribution among subcategories of pages-for example, among all university homepages or all newspaper homepages. Although the connectivity distribution over the entire web is close to a pure power law, we find that the distribution within specific categories is typically unimodal on a log scale, with the location of the mode, and thus the extent of the rich get richer phenomenon, varying across different categories. Similar distributions occur in many other naturally occurring networks, including research paper citations, movie actor collaborations, and United States power grid connections. A simple generative model, incorporating a mixture of preferential and uniform attachment, quantifies the degree to which the rich nodes grow richer, and how new (and poorly connected) nodes can compete. The model accurately accounts for the true connectivity distributions of category-specific web pages, the web as a whole, and other social networks.