• Selected
• Entire List

Abstract:
We present a review of research studies that deal with personalization. We synthesize current knowledge about these areas, and identify issues that we envision will be of interest to researchers working in the management sciences. We take an interdisciplinary approach that spans the areas of economics, marketing, information technology, and operations. We present an overarching framework for personalization that allows us to identify key players in the personalization process, as well as, the key stages of personalization. The framework enables us to examine the strategic role of personalization in the interactions between a firm and other key players in the firms value system. We review extant literature on the strategic behavior of firms, and discuss opportunities for analytical and empirical research in this regard. Next, we examine how a firm can learn a customer's preferences, which is one of the key components of the personalization process. We use a utility-based approach to formalize such preference functions, and to understand how these preference functions could be learnt based on a customers interactions' with a firm. We identify well-established techniques in management sciences that can be gainfully employed in future research on personalization.

Abstract:
Personalization and recommendation systems require knowledge about the users (user profiles), in order to be able to target products, promotions and advertisements. The faster the profiles are learnt, the sooner the site can start benefiting from these systems. We study how a site can learn the profile of a user as quickly as possible as the user traverses the site.

Abstract:
In some classification domains, firms face agents who actively manipulate their information to mislead the firm about their true types so as to avoid unfavorable decisions as a result of the classification. In such domains, firms should take the possibility of applicants' faking behavior into consideration in their decision making. We consider situations where the firm faces agents who can modify instances regardless of their type; unlike prior work, we don't restrict ourselves to those situations where only malicious agents manipulate their data. We show that the firm is never better off when agents have the ability to fake than when they do not. However, surprisingly, a reduction in faking cost does not always hurt the firm, implying that a firm may sometimes prefer an environment in which agents can fake more easily over another in which it is more difficult to fake.

Abstract:
The need to preserve privacy when sharing data across organizations has been recognized as an important issue. In the context of transactional data, privacy is usually preserved by explicitly hiding sensitive information prior to sharing. Often, the data to be shared is stored in a distributed manner by the data owner, where the database is horizontally partitioned to reflect the firms operations in different locations or regions. In such situations, the owner must consider hiding sensitive patterns not only in the consolidated database but also patterns that occur within each partition of the distributed database. We present an Integer Programming (IP) formulation for minimizing data distortion to a distributed database while hiding sensitive patterns. The formulation can become large for distributed databases with multiple partitions and the IP may not be solvable. For such situations, we propose three alternative procedures - Procedure A, Procedure B and Procedure Hybrid - that exploit the distributed nature of the data to decompose the larger problem into a series of smaller problems. We examine the performance of these procedures using computational experiments. The major findings are: i) problems of sizes that cannot be solved to optimality can be solved by these three procedures easily; ii) the differences between the solutions obtained from either Procedure A or Procedure B and the optimal solutions are quite small; and iii) the hybrid procedure, which incorporates in its formulation commonalities between the solutions provided by the other two procedures, is able to obtain solutions that are even closer to the optimal.

data quality, distributed data, accuracy, heuristics, Integer Programming

Abstract:
The focus of this research is to demonstrate how probabilistic models may be used to provide early warnings for bank failures. While prior research in the auditing literature has recognized the applicability of a Bayesian belief revision framework for many audit tasks, empirical evidence has suggested that auditor's cognitive decision processes often violate probability axioms. This study demonstrates that a formal belief revision scheme can be incorporated into an automated system to provide reliable probability estimates for early warning of bank failures. The automated system will examine financial ratios as predictors of a bank's performance, and assess the posterior probability of a bank's financial health. The study examines two different probabilistic models, one that is simpler and makes more assumptions, while the other that is complex but makes fewer assumptions. Both models are able to make accurate predictions while the complex model is well-calibrated in its probability estimates.