* Working papers contributed by faculty members from departments other than information systems.

IS-97-036*	Raghu Garud and S. Kotha Using the Brain as a Metaphor to Model Flexible Productive Units
IS-97-035*	Raghu Garud Commentary: The Process of Relational Contracting
IS-97-034*	Raghu Garud and A. Kumaraswamy Coupling the Technical and Institutional Faces of Janus in Network Industries
IS-97-033*	Raghu Garud, Kumaraswamy and A. Prabhu Networking for Success in Cyberspace
IS-97-032*	Raghu Garud, and A. Kumaraswamy Technological and Organizational Designs to Achieve Economics of Substitution
IS-97-031*	Raghu Garud and S. Jain Technology Embeddedness
IS-97-030*	William H. Starbuck Unlearning Ineffective or Obsolete Technologies
IS-97-029*	William H. Starbuck and John M. Mezias Opening Pandora's Box: Studying the Accuracy of Managers' Perceptions
IS-97-028*	Alan B. Eisner and Zur Shapira Attention Allocation and Managerial Decision Making
IS-97-027*	Sridhar Seshadri, Doron Rotem and Arie Segev Optimal Arrangements of Cartidges in Carousel Type Mass Storage Systems
IS-97-026*	S. Seshadri and D. Rotem The Two Headed Disk: Stochastic Dominance of the Greedy Policy
IS-97-025*	Michael Pinedo and Benjamin P.-C. Yen On the Design and Development of Object-Oriented Scheduling Systems
IS-97-024*	Nicholas Economides, Giuseppe Lopomo and Glenn Woroch Regulatory Pricing Rules to Neutralize Network Dominance
IS-97-023	Ravi Arunkundram and Arun Sundararajan An Economic Analysis of Electronic Secondary Markets: Installed Base, Technology, Durability and Firm Profitability
IS-97-022	Andrew D. Back and Andreas S. Weigend A First Application of Independent Component Analysis to Extracting Structure from Stock Returns
IS-97-021	Raquel Benbunan-Fich Effects of Asynchronous Learning Networks: Results of a Field Experiment Comparing Groups and Individuals
IS-97-020	Shimon Schocken and David Bodoff Framing Effects in Multi-Part and Multi-Stage IT Investment Decisions
IS-97-019	Andreas S. Weigend Data Mining in Finance: Intelligent Information Systems and Computer Intensive Methods for Financial Modeling and Data Analysis
IS-97-018	Abraham Seidmann and Arun Sundararajan Building and Sustaining Inter-Organizational Information Sharing Relationships: The Impact of Interfacing Supply Chain Operations with Marketing Strategy
IS-97-017	Roger M. Stein and Moody's Investors Service A Data Driven Machine Learning Approach to Discovering Rules of Price Behavior in a Financial Market Simulation
IS-97-016	Edward A. Stohr and Yongbeom Kim A Model for Performance Evaluation of Interactive Systems
IS-97-015	Ajit Kambil and David Bodoff Partial Coordination: A Preliminary Evaluation and Failure Analysis
IS-97-014	David Bodoff and Ajit Kambil Pre-Coordination+Post-Coordination=The Case for Partial Coordination
IS-97-013	Blake LeBaron and Andreas S. Weigend A Bootstrap Evaluation of the Effect of Data Splitting on Financial Time Series
IS-97-012	Yongbeom Kim and Edward A. Stohr Software Reuse: Survey and Research Directions
IS-97-011	Jens Timmer and Andreas S. Weigend Modeling Volatility Using State Space Models
IS-97-010	David Bodoff A Re-Unification of Two Competing Models for Document Retrieval
IS-97-009	Henry C. Lucas, Jr. and Valerie Spitler Technology Acceptance and Performance: A Field Study of Broker Workstations
IS-97-008	Gediminas Adomavicius and Alexander Tuzhilin Discovery of Actionable Patterns in Databases: The Action Hierarchy Approach
IS-97-007	Hans Georg Zimmermann, Andreas S. Weigend Exploiting Local Relations as Soft Constraints to Improve Forecasting
IS-97-006	Alex Tuzhilin, Balaji Padmanabhan Unexpectedness as a Measure of Interestingness in Knowledge Discovery
IS-97-005	Mark Choey (Siemens Nixdorfer), Andreas S. Weigend (A.S. Weigend, Y. Abu-Mostafa and A.-P. N. Refenes [eds.]) Nonlinear Trading Models through Sharpe Ratio Maximization (Published in Decision Technologies for Financial Engineering: Proceedings of the Fourth International Conference on Neural Networks in the Capital Markets [NNCM-96]
IS-97-004	Kenneth C. Laudon Extensions to the Theory of Markets and Privacy: Mechanics of Pricing Information
IS-97-003	Raghu Garud, Henry C. Lucas, Jr. Virtual Organizations: What You See May Not Be What You Get
IS-97-002	Shanming Shi (J.P. Morgan), Andreas S. Weigend Taking Time Seriously: Hidden Markov Experts Applied to Financial Engineering
IS-97-001	Chung Yung Simplified Readability Metrics

Raghu Garud
Suresh Kotha
Department of Management
Stern School of Business
New York University
rgarud@stern.nyu.edu

Abstract: Manufacturing flexibility is critical for survival in industries characterized by rapid change and diverse product markets. Although new manufacturing technologies make it possible to accomplish flexibility, their potential remains unrealized by firms whose organizational elements do not possess adaptive capabilities. We use the brain as a metaphor to generate insights on how firms might design flexible production systems. We chose the brain as a metaphor because it is a self-organizing system capable of responding rapidly to a broad range of external stimuli. The brain as a metaphor suggests that flexibility can be enhanced by employing practices that promote distributed processes occurring in a parallel manner. Such practices lie in contrast to those employed by production systems built on scientific management principles that promote localized processes in a sequential manner. By exploring these contrasting modes of operation, we argue that the brain as a metaphor opens up new avenues for theory development related to the design of flexible production systems.

[Return to top of the page]

Raghu Garud
Department of Management
Stern School of Business
New York University
rgarud@stern.nyu.edu

No Abstract

[Return to top of the page]

Raghu Garud
Arun Kumaraswamy
Department of Management
Stern School of Business
New York University
rgarud@stern.nyu.edu

No Abstract

[Return to top of the page]

Raghu Garud
Arun Kumaraswamy
Ajit M. Prabhu
Department of Management
Stern School of Business
New York University
rgarud@stern.nyu.edu

Abstract: Several key technologies are converging to create the emerging cyberspace. We characterize this convergence process as one of cumulative synthesis and suggest that the network mode of organization is the most appropriate for facilitating convergence.

[Return to top of the page]

Raghu Garud
Arun Kumaraswamy
Department of Management
Stern School of Business
New York University
rgarud@stern.nyu.edu

Abstract: Today's industrial landscape is characterized by rapid change and systemic technologies. Rapid change results in ever shorter product life cycles that demand continual innovation from firms. The systemic nature of technologies makes it difficult, if not impossible, for any one firm to manufacture all components of a technological system. We propose that these challenges be met by designing technological systems that have the potential to yield economies of substitution. Additionally, we propose that these economies be realized by adopting the network mode of governance. We examine the network mode at three levels-intrafirm, interfirm, and institutional-to illuminate the inherent tension between cooperation and competition at each level, and to explore the implications of this tension for industrial dynamics.

[Return to top of the page]

Raghu Garud
Sanjay Jain
Department of Management
Stern School of Business
New York University
rgarud@stern.nyu.edu

Abstract: Technological systems are shaped both by forces arising from the technical environment of product markets and those arising from the institutional environment of compatibility standards. We explore how it might be possible for standards to simultaneously enable activities in the technical environment and not constrain them. Such a scenario is possible when the technical environment is not completely embedded in the standards that shape them. We characterize such technological systems as being "just" embedded.

[Return to top of the page]

William H. Starbuck
Department of Management
Stern School of Business
New York University
wstarbuc@stern.nyu.edu

Abstract: Often, before they can learn something new, people have to unlearn what they think they already know. That is, they may have to discover that they should no longer rely on their current beliefs and methods. This paper describes eight viewpoints that can help people to do this.

[Return to top of the page]

William H. Starbuck
John M. Mezias
Department of Management
Stern School of Business
New York University
wstarbuc@stern.nyu.edu

Abstract: Both researchers and managers depend on the accuracy of managers' perceptions. Yet, few studies compare subjective with "objective" data, perhaps because it is very difficult to do well. These difficulties also muddy interpretations of results. On one hand, studies suggest that managers' perceptions may be very inaccurate. On the other hand, the observed errors in managerial perceptions may arise from research methods instead of managers.

Because perceptual data are so significant for both researchers and managers, researchers need to understand both the potential contaminants of perceptual research and the determinants of perceptual errors and biases. This article reviews studies of the accuracies of managers' perceptions, points out hazards in such research, and suggests various ways to improve studies of perceptions. The suggestions encompass improvements in gathering more valid subjective data, locating more appropriate "objective" data, finding appropriate respondents, and using statistical methods that provide accurate and reliable estimates with small samples.

[Return to top of the page]

Zur Shapira
Stern School of Business
New York Univeresity
zshapira@stern.nyu.edu

Abstract: One of the major problems of managerial behavior is the setting of priorities. Time is a scarce resource and managers have to find ways to deal with the multiple tasks that face them. This paper addresses the issue of priority-setting among tasks by managers by proposing analogies from job-shop scheduling theory. We develop a model that views managers employing a combination of rationality and affective judgments with a limited processing capacity.

[Return to top of the page]

Sridhar Seshadri
Doron Rotem
Arie Segev
Stern School of Business
New York University
sseshadr@stern.nyu.edu

Abstract: Optimal arrangements of cartridges and file partitioning schemes are examined in carousel type mass storage systems using Markov decision theory. It is shown that the Organ-Pipe Arrangement is optimal under different storage configurations for both the anticipatory as well as the non-anticipatory versions of the problem. When requests arrive as per an arbitrary renewal process this arrangement is also shown to minimize the mean queueing delay and the time spent in the system by the requests.

[Return to top of the page]

Sridhar Seshadri
Doron Rotem
Stern School of Business
New York University
sseshadr@steren.nyu.edu

Abstract: In his paper "Should the two-headed disk be greedy? - Yes, it should" Hofri defined a "greedy policy" as follows. Assuming that the range of disk addresses is [0,1], request at location x is served by the closet arm while the other arm jockeys to a new position, z, where z = (1/3)x or z = 2/3 + x/3 depending on whether x is larger or smaller than ½. Hofri proved that his policy minimizes the expected seek distance for uniform request probabilities and conjectured that it stochastically dominates every other policy. Stochastic dominance is of practical importance in this context as it guarantees that a policy that optimizes expected seek distance also guarantees optimal seek time. The main result of this paper is a proof of Hofri's conjecture. The paper contains two proofs, the first establishes the conjecture, and the second shows that if the seek distance is stochastically minimized under a repositioning policy, then the policy must be Hofri's greedy policy and the request distribution must be uniform.

[Return to top of the page]

Michael Pinedo
Stern School of Business
New York University
mpinedo@stern.nyu.edu

Benjamin P.-C. Yen

Abstract: In this paper, we describe the architecture of an object-oriented scheduling system. First, a mathematical framework is presented that is based on set theory and graph theory. Then a number of basic as well as more specialized methods are defined which can be applied on the entities of any decision support system. The principal objects of a scheduling system are defined, as well as the methods specifically designed for the manipulation of the schedules. The object base design, the schedule generator design and the user interface design are then discussed in detail.

[Return to top of the page]

Full Paper (Adobe PDF format)

Glenn Woroch
Department of Economics
University of California, Berkeley
glenn@econ.berkeley.edu

Nicholas Economides
Giuseppe Lopomo
Stern School of Business
New York University
neconomi@stern.nyu.edu, glopomo@stern.nyuu.edu

ABSTRACT: This paper evaluates the effectiveness of several pricing rules intended to promote entry into a network industry dominated by an incumbent carrier. Drawing on the work of Cournot and Hotelling, we develop a model of competition between two interconnected networks. In a symmetric equilibrium, the price of cross-network calls exceeds the price of internal calls. This "calling circle discount" tends to "tip" the industry to a monopoly equilibrium as would a network externality. By equalizing charges for terminating calls, reciprocity eliminates differences between internal and cross-network prices and makes monopoly less likely. Imputation counteracts an incentive by the dominant network to "price squeeze" a rival by eliminating differences in the wholesale price of termination and the implicit price for internal use. By increasing profits of rival networks and increasing their subscribers' surplus, imputation supports additional entry. Finally, an unbundling rule reduces termination fees charged by a dominant network that was engaging in pure bundling. Again, entry will be facilitated as rival networks offer potential subscribers a more attractive rate schedule.

JEL classification: L1, D4
Key words: two-way networks, interconnection, reciprocity, imputation, unbundling

[Return to top of the page]

Full Paper (Adobe PDF and Postscript Format)

Andrew D. Back

Andreas S. Wiegend
Department of Information Systems
Stern School of Business
New York University
email: aweigend@stern.nyu.edu

Abstract: This paper discusses the application of a modern signal processing technique known as independent component analysis (ICA) or blind source separation to multivariate financial time series such as a portfolio of stocks. The key idea of ICA is to linearly map the observed multivariate time series into a new space of statistically independent components (ICs). This can be viewed as a factorization of the portfolio since joint probabilities become simple products in the coordinate system of the ICs.

We apply ICA to three years of daily returns of the 28 largest Japanese stocks and compare the results with those obtained using principal component analysis. The results indicate that the estimated ICs fall into two categories, (1) infrequent but large shocks (responsible for the major changes in the stock prices), and (2) frequent smaller fluctuations (contributing little to the overall level of the stocks). We show that the overall stock price can be reconstructed surprisingly well by using a small number of thresholded weighted ICs. In contrast, when using shocks derived from principal components instead of independent components, the reconstructed price is less similar to the original one. Independent component analysis is a potentially powerful method of analyzing and understanding driving mechanisms in financial markets. There are further promising applications to risk management since ICA focuses on higher order statistics.

[Return to top of the page]

Raquel Benbunan-Fich
Department of Information Systems
Stern School of Business
New York University
email: raquel@stern.nyu.edu

Abstract: An Asynchronous Learning Network (ALN) is a Computer-Mediated Communication System designed to support "anytime/anywhere" interaction among students and between students and instructors. A field experiment compared groups and individuals solving an ethical case scenario, with and without an ALN, to determine the separate and joint effects of communication medium and teamwork. Undergraduate students in Computers and Society analyzed the case as an assignment in the course. Dependent variables include quality of the reports, learning as measured by similar cases on the final exam, and subjective perceptions of learning.

The results indicate that working in a group, instead of alone, tends to increase motivation, perception of learning and solution satisfaction. Individuals working online produced higher quality reports on the ethics scenario than individuals working manually, and computer-supported groups produced the longest reports, while individuals working manually produced the shortest reports. Regarding group conditions, manual teams reported significantly higher levels of process satisfaction, perception of process structure and perception of discussion quality than teams supported by an asynchronous communication medium. However, computer-supported groups reported the highest levels of perceived learning. Finally, perception of collaborative learning does not seem to be affected by the use of the medium; both supported and unsupported groups perceived about the same levels of collaborative learning.

[Return to top of the page]

David Bodoff
Department of Information Systems
Stern School of Business
New York University
dbodoff@stern.nyu.edu

Abstract: Normative decision-theories specify how managerial decision-makers ought to behave when facing complex decisions. Prospect Theory, on the other hand, describes "framing effects" -- psychological factors which affect decision-makers when they confront complex decisions in actual practice. These psychological framing effects are prominent in many complex IT investment decisions. This paper applies Prospect Theory to two kinds of complex decisions, i.e. multi-stage and multi-part IT investment decisions. The theory predicts that psychological framing effects will lead IT managers to normatively undesirable behavior. An experiment is conducted which provides some support for Prospect Theory as it is applied to IT investment decision.

[Return to top of the page]

Full Paper (Adobe PDF and Postscript Format)

Andreas S. Weigend
Department of Information Systems
Stern School of Business
New York University
email: aweigend@stern.nyu.edu

Abstract: The term "data mining" refers to new methods for the intelligent analysis of large data sets. These methods have emerged from several historically disjoint fields, such as applied statistics, information systems, machine learning, data engineering, artificial intelligence, and knowledge discovery. One of the most enticing application areas of these emerging technologies is finance, becoming more amenable to data-driven modeling as large sets of financial data become available.

This Working Paper describes a new graduate course (B20.3355) at Stern that provides links among Information Systems, Statistics, and Finance. It provides the relevant background knowledge from the reference disciplines, presents the foundations of data mining methods as well as their computer implementations, and applies them to current problems in finance, including the building and evaluating of trading models and the managing of risk.

Besides the two lectures each week (that introduce the new approaches and discuss their advantages and disadvantages), there are two additional elements to the didactics of this course:
- Weekly hands-on homework assignments that help students gain familiarity with the corresponding tools, and a
- A major project that deeply explores one or two of the approaches from class on a real-world problem.

Each project is carried out in conjunction with a major Wall Street firm. This is a unique opportunity for the student to apply one of the concepts introduced in the course to a genuine, current business situation. Besides the in-depth learning and understanding of the technical aspects, this provides insights into the challenges for data mining groups, as well as experience and connections for future jobs. These projects are carried out in teams, improving the students' communication and team skills.

[Return to top of the page]

Roger M. Stein
Department of Information Systems
Stern School of Business
New York University
email: rstein@stern.nyu.edu

Abstract: The field of agent-based simulation of financial markets has grown considerably in the last decade. However, the interpretation of simulation results has received far less attention. Typically, the results of a large number of simulations are reduced to one or two summary statistics, such as sample moments. While such summarization is useful, it overlooks a vast amount of additional information that might be gleaned by examining patterns of behavior that emerge at lower levels. In this paper we propose an approach to interpreting simulation results that involves the use of so-called data mining techniques to identify the rules of behavior that govern an underlying system. We demonstrate the approach by using data from a single run of an order market simulation to derive rules about the behavior of prices in that simulation.

[Return to top of the page]

Edward A. Stohr
Yongbeom Kim
Department of Information Systems
Stern School of Business
New York University
email: estohr@stern.nyu.edu

Abstract: We describe a quantitative model for the performance evaluation of interactive computer systems. The approach involves the development of an "interaction graph" or state transition diagram to describe the user-machine interaction. Given numerical data on transition times and probabilities, the model can be used to perform sensitivity analyses of changes in system parameters and user behavior. To illustrate the model, we use empirical data from field and laboratory experiments designed to compare a prototype natural language query system with a formal (relational) query system. The general approach is applicable in a broad range of other contexts including bibliographic retrieval and the analysis of web-log data. It should be of interest to both system developers and potential users of these systems.

[Return to top of the page]

Ajit Kambil
David Bodoff
Department of Information Systems
Stern School of Business
New York University
email: akambil@stern.nyu.edu, dbodoff@stern.nyu.edu

Abstract: Partial coordination is a new method for cataloging documents for subject access. It is especially designed to enhance the precision of document searches in online environments. This paper reports a preliminary evaluation of partial coordination which shows promising results compared with full text retrieval. We also report the difficulties in empirically evaluating the effectiveness of automatic full-text retrieval in contrast to mixed methods such as partial coordination which combine human cataloging with computerized retrieval. Based on our study we propose research in this area will substantially benefit from a common framework for failure analysis and a common data set. This will allow information retrieval researchers adapting "library style" cataloging to large electronic document collections, as well as those developing automated or mixed methods, to directly compare their proposals for indexing and retrieval. This paper concludes by suggesting guidelines for constructing such a testbed.

[Return to top of the page]

David Bodoff
Ajit Kambil
Department of Information Systems
Stern School of Business
New York University
email: akambil@stern.nyu.edu, dbodoff@stern.nyu.edu

Abstract: The introduction of computerized post-coordination has solved many of the problems of pre-coordinated subject access. However, the adoption of computerized post-coordination results in the loss of some pre-coordination benefits. Specifically, the effect of hiding terms within the context of others is lost in post-coordination which gives lead status to every document term. This results in spurious matches of terms out of context. Library patrons and Internet searchers are increasingly dissatisfied with subject access performance, in part because of unmanageably large retrieval sets. The need to enhance precision and limit the size of retrieval sets motivates this work which proposes partial coordination, an approach which incorporates the advantages of computer search with the ability of pre-coordination to limit spurious partial matches and thereby enhance precision.

[Return to top of the page]

Full Paper (Adobe PDF and Postscript Format)

Blake Le Baron
Department of Economics
University of Wisconsin
Madison, WI 53713
blebaron@facstaff.wisc.edu

Andreas S. Weigend
Department of Information Systems
Stern School of Business
New York University
aweigend@stern.nyu.edu

July 1997

ABSTRACT: This article exposes problems of the commonly used technique of splitting the available data into training, validation, and test sets that are held fixed, warns about drawing too strong conclusions from such static splits, and shows potential pitfalls of ignoring variability across splits. Using a bootstrap or resampling method, we compare the uncertainty in the solution stemming from the data splitting with neural network specific uncertainties (parameter initialization, choice of Number of hidden units, etc.). We present two results on data from the New York Stock Exchange. First, the variation due to different resamplings is significantly larger than the variation due to different network conditions. This result implies that it is important to not over-interpret a model (or an ensemble of models) estimated on one specific split of the data. Second, on each split, the neural network solution with early stopping is very close to a linear model; no significant nonlinearities are extracted.

Keywords: Model evaluation, model uncertainty, bookstrap, resampling, financial forecasting, timeseries prediction, linear bias of early stopping, superposition of forecasts, model merging.

Data: Dow Jones Industial Average, 1962-1987. Volume from New York Stock Exchange, 1962-1987. Data used in this article is available from the Web sites of the authors.

[Return to top of the page]

Yongbeom Kim
Edward Stohr
Department of Information Systems
Stern School of Business
New York University
estohr@stern.nyu.edu

Abstract: To be furbished

[Return to top of the page]

Full Paper (Adobe PDF and Postscript Format)

Jens Timmer
Fakultät für Physik
Universität Freiburg
Freiburg, Germany
jeti@fdm.uni-freiburg.de

Andreas S. Weigend
Department of Information Systems
Stern School of Business
New York University
aweigend@stern.nyu.edu

July 1997

ABSTRACT: The powerful framework of state space modeling is only useful if the time series contains observational noise. After discussing signatures of observational noise, we found no evidence for observational noise on spot price series and on returns. However, on squared returns a significant amount of observational noise can be detected. This allows us to estimate a full state space model that explicitly includes observational noise. We compare its predictive performance to standard AR models that ignore the distinction between observational and dynamical noise and underestimate the parameters. Applications include the estimation of volatilities, the detection of mispricings, the computation of value at risk, and the construction of trading models.

[Return to top of the page]

David Bodoff
Department of Information Systems
Stern School of Business
New York University
dbodoff@stern.nyu.edu

June 1997

ABSTRACT: Two competing approaches for document retrieval were first identified by Robertson et al. (Robertson, Maron et al., 1982) for probabilistic retrieval. We point out the corresponding two competing approaches for the Vector Space Model. In both the probabilistic and Vector Space models, only one of the two competing approaches has received significant research attention, because of the unavailability of sufficient data to implement the second approach. Because it is now feasible to collect vast amounts of feedback data from users, both approaches are now possible. We therefore re-visit the question of a unification of both approaches, for both probabilistic and Vector Space models. This uification of approaches differs from that originally proposed in (Robertson, Maron et al, 1982), and offers unique advantages. Preliminary results of a simulation experiment are reported, and an outline is provided of an ongoing field study.

[Return to top of the page]

Henry C. Lucas, Jr.
Department of Information Systems
Stern School of Business
New York University
hlucas@stern.nyu.edu

Valerie Spitler
Department of Information Systems
Stern School of Business
New York University
vspitler@stern.nyu.edu

June 1997

ABSTRACT: We develop a model to predict 1) the use of a multifunctional, broker workstation with a windowed interface and 2) the relationship between workstation use and performance. Brokers and sales assistants in the private client group of a major investment bank use this workstation as an integral part of their jobs. Our model explains some of the variance in their usage, intended usage and performance, but the variables that are most salient in the model differ between brokers and sales assistants. There is evidence that low performing brokers use the workstation more than higher performing brokers; the results also suggest that a different type of training may be needed for sophisticated workstations for professionals than for clerical personnel learning to use transactions processing systems. We believe it is important to understand the acceptance of technology and the relationship between system use and performance if firms are to obtain a return from investing in information technology.

This research was sponsored in part by National Science Foundation Grant number IRI-9200205.

[Return to top of the page]

Alexander Tuzhilin
Department of Information Systems
Stern School of Business
New York University
atuzhili@stern.nyu.edu

June 1997

ABSTRACT: To be furnished

[Return to top of the page]

Full Paper (Adobe PDF and Postscript Format)

Hans Georg Zimmermann

Andreas S. Weigend
Department of Information Systems
Stern School of Business
New York University
e-mail: aweigend@stern.nyu.edu

georg.zimmermann@mchp.siemens.de

ABSTRACT: Predictive models for financial data are often based on a large number of possible inputs that are potentially nonlinearly combined to approximate a target. The target is typically a single number, such as a return of an asset. This paper introduces a new architecture: On the output side, we predict dynamical variables such as first derivatives and curvatures on different time spans. These are subsequently combined in an interaction output layer to several estimates of the prediction of interest. Those estimates are then averaged to yield the final prediction. On the input side, we are including a new internal preprocessing layer connected with a diagonal matrix of positive weights to a layer of squashing functions. These weights adapt for each input individually to squash outliers. We apply these ideas to the real world example of the daily predictions of the German stock index DAX (Deutscher Aktien Index), and compare the results to a state-of-the-art network with a single output. It turns out that the new six layer architecture is more stable in training. We believe that this is due to the larger amount of information flowing back from the outputs in the backward pass for the parameter adaptation, as well as to the constraints of predicting first and second derivatives, approximating relevant variables in the dynamics.

Keywords: Interaction layer, Feature layer, Preprocessing Layer, Dynamical systems variables, Momentum, Forces, Stability, Overfitting

Data set used: German Stock Index (DAX, Deutscher Aktien Index)

[Return to top of the page]

Department of Information Systems
Stern School of Business
New York University
e-mail: atuzhili@stern.nyu.edu, bpadmana@stern.nyu.edu

ABSTRACT: to be furnished

Full Paper (Adobe PDF and Postscript Format)

Mark Choey (Siemens Nixdorfer)

Andreas S. Weigend
Department of Information Systems
Stern School of Business
New York University
e-mail: aweigend@stern.nyu.edu

ABSTRACT: While many trading strategies are based on price prediction, traders in financial markets are typically interested in risk-adjusted performance such as the Sharpe Ratio, rather than price predictions themselves. This paper introduces an approach to generate a nonlinear strategy that explicitly maximizes the Sharpe Ratio. It is expressed as a neural network model that outputs the position size between a risky and a risk-free asset. The iterative parameter update rules are derived and compared to alternative approaches. The resulting trading strategy is evaluated and analyzed on both computer-generated data and real world data, the daily German Stock Index (DAX). The trading based on Sharpe Ratio maximization compares favorably to both profit optimization and to probability matching (through cross-entropy optimization). The results presented show that the goal of optimizing out-of-sample risk adjusted profit can be achieved with this nonlinear approach.

Keywords: Risk, Trading Strategies, Allocation, Sharpe Ratio, Risk-reward ratio, Profit optimization, Cross entropy, Probability matching, Decision technologies.

Data set used: German Stock Index (DAX, Deutscher Aktien Index)

[Return to top of the page]

Kenneth C. Laudon
Department of Information Systems
Stern School of Business
New York University
e-mail: klaudon@stern.nyu.edu

ABSTRACT: This is the second paper in a series concerned with the theory of markets and privacy. The first paper showed how market theory and privacy theory and practice are related. It is argued there that the current crisis in privacy is a result of market failure rather than technological progress. If markets would be allowed to operate, and individuals allowed to sell their personal information to buyers, then the crisis in privacy would be ameliorated because individuals would be given control over the fate of their personal information. Many questions are raised by this approach and will be addressed in following papers. One question concerns establishing the price of personal information.

This paper examines the buyer point of view. The question is: how much should a purchaser pay for personal information. We review four related models for establishing the purchase offer price: discounted cash flow with no learning, discounted cash flow with learning, discounted cash flow with "learning" and "loss avoidance" effects, and options models. We conclude with a consideration of market strategies and the issue of co-specialized information assets. A following paper reviews different models for understanding the seller point of view where the question is: how much should I charge for my personal information.

[Return to top of the page]

Raghu Garud
Department of Management
rgarud@stern.nyu.edu

Henry C. Lucas, Jr.
Department of Information Systems
hlucas@stern.nyu.edu

Stern School of Business
New York University

ABSTRACT: Virtual organizations are new organizational forms comprising a set of network transactions that differ from those found in markets and hierarchies. This paper explores the nature of these network transactions through an in-depth study of a virtual firm. The virtual organization is characterized by constant organizing through virtual teams and alliances, a unique management culture and a set of norms, information and knowledge sharing enabled by information technology, and employee self-governance. The organization gains from a culture of fast-response and efficiency while employees are trusted to exercise discretion and take initiatives.

[Return to top of the page]

Full Paper (Adobe PDF and Postscript Format)

Shanming Shi
J.P. Morgan

Andreas S. Weigend
Department of Information Systems
Stern School of Business
New York University
e-mail: aweigend@stern.nyu.edu
http://www.stern.nyu.edu/~aweigend

ABSTRACT: Most traditional time series models are global models based on local time information: they assume that the state can be fully and locally (in time) characterized with a finite embedding space. Prediction then amounts to simple regression. Unfortunately, there are many situations in which simple regression is not sufficient to model the temporal structure in a time series. We here introduce an architecture that we call Hidden Markov Experts. It is based on Hidden Markov Models used in speech recognition research. By introducing the concept of hidden states, Hidden Markov experts model time dependency of time series explicitly as a first-order Markov model with transitions between these hidden states. Within each state, local models are applied to estimate the probability density, which can be linear or nonlinear depending on the situation. This paper first discusses the statistical framework and the learning algorithm of Hidden Markov experts, then applies them to daily S&P500 data and to high frequency currency exchange rate data. The Hidden Markov Experts have better profit than the linear and nonlinear global models. The volatilities of the time series can be characterized by the hidden states.

Keywords: Regime Switching, Hidden States, Probability Density Prediction, Non-constant Transition Probabilities, EM Algorithm, Risk Estimation Data sets used: High-frequency DEM/USD exchange rates. Daily S&P500.

[Return to top of the page]

Chung Yung
Courant Institutee
New York University
yung@edgar.stern.nyu.edu

ABSTRACT: This paper describes a new approach to measuring the complexity of software systems with considering their readability. Readability Metrics were proposed by Chung and Yung in 1990. Readability Metrics have been outstanding among the existing software complexity metrics for taking nonphysical software attributes, like readability, into considerations. The applications of Readability Metrics are good in indicating the additional efforts required for less readable software systems, and help in keeping the software systems maintainable. However, the numerous metrics and the complicated formulas in the family usually make it tedious to apply Readability Metrics to large scale software systems. In this paper, we propose a simplified approach to Readability Metrics. We reduce the number of required measures and keep the considerations on software readability. We introduce our Readability model in a more formal way. The Readability Metrics preprocesses algorithm is developed with compilers front-end techniques. The experiment results show that this simplified approach has good predictive power in measuring software complexity with software readability, in addition to its ease of applying. The applications of Readability Metrics indicate the readability of software systems and help in keeping the source code readable and maintainable.

[Return to top of the page]