http://hdl.handle.net/2451/31634 2026-04-11T15:08:52Z 2026-04-11T15:08:52Z Clark, Jessica Paiement, Jean Francois Provost, Foster http://hdl.handle.net/2451/36731 2016-10-24T20:34:08Z 2016-10-24T00:00:00Z

Title: Who's Watching TV? Authors: Clark, Jessica; Paiement, Jean Francois; Provost, Foster Abstract: TV viewership data available at the individual set-top box level has enabled new methods for estimating the demographics of shows' audiences, but it is impossible to tell with certainty which household members are watching TV in multi-person households. We address this problem through four main contributions. First, we develop a novel method for estimating the likelihood that each individual in a multi-person household is watching. Second, we derive a set of tasks at which models must succeed in order to demonstrate that they have solved the core problem, since there are no ground-truth labels. Third, we evaluate our new method as well as two current state-of-the-art heuristic methods. Fourth, we conduct some example analyses of viewership in the context of living with others. Our solution has implications for advertisers, researchers who seek better understanding TV viewership, and anyone using data generated by shared devices or accounts. A major TV provider is planning on deploying this method for use in their TV ad-targeting system. No personally identifiable information (PII) was gathered or used in conducting this study. To the extent any data was analyzed, it was anonymous and/or aggregated data, consistent with the carrier's privacy policy.

2016-10-24T00:00:00Z Clark, Jessica Provost, Foster http://hdl.handle.net/2451/36320 2016-09-29T20:01:52Z 2016-09-29T00:00:00Z

Title: Matrix-Factorization-Based Dimensionality Reduction in the Predictive Modeling Process: A Design Science Perspective Authors: Clark, Jessica; Provost, Foster Abstract: Dimensionality Reduction (DR) is frequently employed in the predictive modeling process with the goal of improving the generalization performance of models. This paper takes a design science perspective on DR.We treat it as an important business analytics artifact and investigate its utility in the context of binary classification, with the goal of understanding its proper use and thus improving predictive modeling research and practice. Despite DR's popularity, we show that many published studies fail to undertake the necessary comparison to establish that it actually improves performance. We then conduct an experimental comparison between binary classification with and without matrix-factorization-based DR as a preprocessing step on the features. In particular, we investigate DR in the context of supervised complexity control. These experiments utilize three classifiers and three matrix-factorization based DR techniques, and measure performance on a total of 26 classification tasks. We find that DR is generally not beneficial for binary classification. Specifically, the more difficult the problem, the more DR is able to improve performance (but it diminishes easier problems' performance). However, this relationship depends on complexity control: DR's benefit is actually eliminated completely when state-of-the-art methods are used for complexity control. The wide variety of experimental conditions allows us to dig more deeply into when and why the different forms of complexity control are useful. We find that L2-regularized logistic regression models trained on the original feature set have the best performance in general. The relative benefit provided by DR is increased when using a classifier that incorporates feature selection; unfortunately, the performance of these models, even with DR, is lower in general. We compare three matrix-factorization-based DR algorithms and nd that none does better than using the full feature set, but of the three, SVD has the best performance. The results in this paper should be broadly useful for researchers and industry practitioners who work in applied data science. In particular, they emphasize the design science principle that adding design elements to the predictive modeling process should be done with attention to whether they add value.

2016-09-29T00:00:00Z Panniello, Umberto Gorgoglione, Michele Tuzhilin, Alexander http://hdl.handle.net/2451/34278 2015-10-05T17:55:38Z 2015-10-05T00:00:00Z

Title: In CARSWe Trust: How Context-Aware Recommendations Affect Customers’ Trust And Other Business Performance Measures Of Recommender Systems Authors: Panniello, Umberto; Gorgoglione, Michele; Tuzhilin, Alexander Abstract: Most of the work on Context-Aware Recommender Systems (CARSes) has focused on demonstrating that the contextual information leads to more accurate recommendations and on developing efficient recommendation algorithms utilizing this additional contextual information. Little work has been done, however, on studying how much the contextual information affects purchasing behavior and trust of customers. In this paper, we study how including context in recommendations affects customers’ trust, sales and other crucial business-related performance measures. To do this, we performed a live controlled experiment with real customers of a commercial European online publisher. We delivered content-based recommendations and context-aware recommendations to two groups of customers and to a control group. We measured the recommendations’ accuracy and diversification, how much customers spent purchasing products during the experiment, quantity and price of their purchases and the customers’ level of trust. We aim at demonstrating that accuracy and diversification have only limited direct effect on customers’ purchasing behavior, but they affect trust which drives the customer purchasing behavior. We also want to prove that CARSes can increase both recommendations’ accuracy and diversification compared to other recommendation engines. This means that including contextual information in recommendations not only increases accuracy, as was demonstrated in previous studies, but it is crucial for improving trust which, in turn, can affect other business-related performance measures, such as company’s sales.

2015-10-05T00:00:00Z Bauman, Konstantin Tuzhilin, Alexander Zaczynski, Ryan http://hdl.handle.net/2451/34226 2015-09-01T14:48:48Z 2015-09-01T00:00:00Z

Title: Virtual Power Outage Detection Using Social Sensors Authors: Bauman, Konstantin; Tuzhilin, Alexander; Zaczynski, Ryan Abstract: In this paper we describe a novel approach to detecting power outages that utilizes social media platform users as “social sensors” for virtual detection of power outages. We present the underlying methodology based on analyzing Twitter and other social media data that detects bursts in tweets related to the power outages. The proposed methodology was implemented and deployed by a major company in the area of enterprise solutions for social media aggregation for the electrical utility industry as a part of their comprehensive social engagement platform. It was also field tested on the Twitter users in an industrial setting and performed well during these tests.

2015-09-01T00:00:00Z