http://hdl.handle.net/2451/14813 Thu, 02 Apr 2026 04:24:02 GMT 2026-04-02T04:24:02Z http://hdl.handle.net/2451/31553 Title: Data Science and Prediction Authors: Dhar, Vasant Abstract: The world's data is growing more than 40% annually. Coupled with exponentially growing computing horsepower, this provides us with unprecedented basis for 'learning' useful things from the data through statistical induction without material human intervention and acting on them. Philosophers have long debated the merits and demerits of induction as a scientific method, the latter being that conclusions are not guaranteed to be certain and that multiple and numerous models can be conjured to explain the observed data. I propose that 'big data' brings a new and important perspective to these problems in that it greatly ameliorates historical concerns about induction, especially if our primary objective is prediction as opposed to causal model identification. Equally significantly, it propels us into an era of automated decision making, where computers will make the bulk of decisions because it is infeasible or more costly for humans to do so. In this paper, I describe how scale, integration and most importantly, prediction will be distinguishing hallmarks in this coming era of Data Science.' In this brief monograph, I define this newly emerging field from business and research perspectives. Tue, 29 May 2012 20:30:46 GMT http://hdl.handle.net/2451/31553 2012-05-29T20:30:46Z http://hdl.handle.net/2451/31303 Title: Comparative Effectiveness for Oral Anti-diabetic Treatments among Newly Diagnosed Type-2 Diabetics: Machine Learning Applied to a Large-Scale Claims Dataset Authors: Maguire, Jon; Dhar, Vasant Abstract: In this paper, we demonstrate how the US healthcare system can provide increased benefits per unit of spend, and earlier identification of and intervention in chronic diseases through better predictive data-based analytics applied to the increasingly available troves of healthcare claims data. Specifically, we demonstrate the effectiveness of data mining by applying machine learning methods to large-scale medical and pharmacy claims data for roughly 70,000 patients over six years on newly diagnosed with type-2 diabetes, a common disease in the US costing billions to treat. This analysis reveals important differences in cost and quality among the disease's common treatments some of which have been published in the American Diabetes Association, and others that are regarded as tentative or have not been considered at all. The study demonstrates the potential for using large scale data mining for better understanding other major diseases including coronary problems and cancers and for focusing further inquiry in these areas. Mon, 07 Nov 2011 21:18:23 GMT http://hdl.handle.net/2451/31303 2011-11-07T21:18:23Z http://hdl.handle.net/2451/31279 Title: Geo-Social Targeting for Privacy-friendly Mobile Advertising: Position Paper Authors: Provost, Foster Abstract: This position paper is about methods for effective, privacy-friendly mobile advertising. Specifically, we propose a new social-targeting design for using consumer location data from mobile devices (smart phones, smart pads, laptops, etc.) to target advertisements in a manner that is both effective and privacy friendly. This paper introduces the main concepts, provides theoretical background and ties to the literature, presents the design itself, and discusses applications. Tue, 18 Oct 2011 13:06:45 GMT http://hdl.handle.net/2451/31279 2011-10-18T13:06:45Z http://hdl.handle.net/2451/31253 Title: Pseudo-social network targeting from consumer transaction data Authors: Martens, David; Provost, Foster Abstract: This design science paper presents a method for targeting consumers based on a 'pseudo-social network' (PSN): consumers are linked if they transfer money to the same entities. A marketer can target those individuals that are strongly connected to key individuals. We present the PSN design and a large-scale empirical study using data from a major bank. For two different product offerings, consumers that are close to existing customers in the PSN have significantly higher take rates than the 'most likely' candidates identified by state-of-the-art socio-demographic (SD) predictive modeling. Interestingly, the PSN targeting only does better for the closest neighbors. However, the different models capture different information: combining the two does significantly better than either alone. The results demonstrate that social targeting can be applied broadly, to settings where the network among consumers is unlikely to be a true social network, but nonetheless captures inherent similarity. Mon, 26 Sep 2011 14:45:06 GMT http://hdl.handle.net/2451/31253 2011-09-26T14:45:06Z