Discovery of financial fraud has profound social consequences. Loss of stockholder value, bankruptcy, and loss of confidence in the professional audit firms have resulted from failure to detect financial fraud. Previous studies that have attempted to discover fraud patterns from publicly available information have achieved only moderate levels of success. This study explores the capabilities of recently developed statistical learning and data mining methods in an attempt to advance fraud discovery performance to levels that have potential for proactive discovery or mitigation of financial fraud. The partially adaptive methods we test have achieved success in a number of complex problem domains and are easily interpretable. Ensemble methods, which combine predictions from multiple models via boosting, bagging, or related approaches, have emerged as among the most powerful data mining and machine learning methods. Our study includes random forests, stochastic gradient boosting, and rule ensembles. The results for ensemble models show marked improvement over past efforts, with accuracy approaching levels of practical potential. In particular, rule ensembles do so while maintaining a degree of interpretability absent in the other ensemble methods.

This paper identifies and defines standards required for successful e-commerce (EC) platforms. It presents an ontology of standards and applications that can be used to evaluate any EC platform. Current platforms are limited in usefulness and development economy by lack of standards. We describe how the development and use of such standards could help improve EC platforms. We examine the current and emerging platforms and technologies in relation to the proposed ontology including electronic data interchange, e-procurement systems, business-to-business hubs, company WeWeb b sites, electronic business eXtensible markup language, Web services, the semantic Web, and universal business language. We conclude that no individual platform or technology provides complete support for every level of the ontology, and we suggest that combinations of features from various platforms and technologies may be of most value. We conclude that research is needed in three areas: the discovery phase of EC, participant incentives for standardization, and methods of combining useful aspects from different standards bodies.

Model checking is a promising technique for the verification of complex software systems. As the use of the Internet for conducting e-business extends the reach of many organizations, well-designed software becomes the foundation of reliable implementation of e-business processes. These distributed, electronic methods of conducting transactions place reliance on the control structures embedded in the transaction processes. Deficiencies in control structures of processes that support e-business can lead to loss of physical assets, digital assets, money, and consumer confidence. Yet, assessing the reliability of e-business processes is complex and time-consuming. This paper explicates how model-checking technology can aid in the design and assurance of e-business processes in complex digital environments. Specifically, we demonstrate how model checking can be used to verify e-business requirements concerning money atomicity, goods atomicity, valid receipt, and communication-link failure. These requirements are fundamental to many e-business applications. Model checking can be used to test a broad range of systems requirements-not only for system designers, but also for auditors and security specialists. Systems that are examined by auditors need to have adequate controls built in prior to implementation and will need adequate auditing after implementation to ensure that none of the processes have been corrupted. Model checkers may also provide value in examining the processes of highly integrated applications as found in enterprise resource planning systems.

Time-series analysis is important to a wide range of disciplines transcending both the physical and social sciences. Statistical models have sound theoretical bases and have been successfully used in a number of problem domains. More recently, machine-learning models such as neural networks have been suggested as offering potential for time-series analysis. Results of neural network empirical testing have thus far been mixed. This paper proposes melding useful parameters from the statistical ARIMA model with neural networks of two types: multilevel perceptrons (MLPs) and radial basis functions (RBFs). Tests are run on a range of time-series problems that exhibit many common patterns encountered by analysts. The results suggest that hybrids of the type proposed may yield better outcomes than either model by itself.

While some efforts have been made to formalize the choice of parameters for artificial neural networks (multi-level perceptrons) using the backpropagation (BP) algorithm, studies reported in the literature to date appear to be dominated by use of heuristic methods of design. The objective of this paper is to test the comparative performance of neural networks with parameters determined by genetic algorithms (GAs). We have attempted to improve the quality of this type of research by comparing several models across seven real problem domains, including both regression and classification problems. Performance is benchmarked against neural networks whose parameters are selected using heuristic methods, as well as several widely used statistical models. For the problem domains tested, GA-assisted design yields models that consistently perform as well or better than heuristic methods and the statistical models.

Neural networks whose architecture is determined by genetic algorithms outperform autoregressive integrated moving average forecasting models in six different time series examples. Refinements to the autoregressive integrated moving average model improve forecasting performance over standard ordinary least squares estimation by 8% to 13%. In contrast, neural networks achieve dramatic improvements of 10% to 40%. Additionally, neural networks give evidence of detecting patterns in data which remain hidden to the autoregression and moving average models. The consequent forecasting potential of neural networks makes them a very promising addition to the variety of techniques and methodologies used to anticipate future movements in time series.

Ever since the initial planning for the 1997 Utah legislative session, neural-network forecasting techniques have provided valuable insights for analysts forecasting tax revenues. These revenue estimates are critically important since agency budgets, support for education, and improvements to infrastructure all depend on their accuracy. Underforecasting generates windfalls that concern taxpayers, whereas overforecasting produces budget shortfalls that cause inadequately funded commitments. The pattern finding ability of neural networks gives insightful and alternative views of the seasonal and cyclical components commonly found in economic time series data. Two applications of neural networks to revenue forecasting clearly demonstrate how these models complement traditional time series techniques. In the first, preoccupation with a potential downturn in the economy distracts analysis based on traditional time series methods so that it overlooks an emerging new phenomenon in the data. In this case, neural networks identify the new pattern that then allows modification of the time series models and finally gives more accurate forecasts. In the second application, data structure found by traditional statistical tools allows analysts to provide neural networks with important information that the networks then use to create more accurate models. In summary, for the Utah revenue outlook, the insights that result from a portfolio of forecasts that includes neural networks exceeds the understanding generated from strictly statistical forecasting techniques. In this case, the synergy clearly results in the whole of the portfolio of forecasts being more accurate than the sum of the individual parts.

A major complexity in the development of many computerized information systems is the lack of an integrated framework and analysis structure to consolidate disparate types of information. Successful information system development in complex environments requires an approach that supports development of systems across all information areas and uses the same model for analysis, specification, prototyping, design and implementation. This paper presents a robust methodology for dealing with these needs. The method has a sound theoretical basis, yet is straightforward to understand and use. An application from computerized manufacturing is used to illustrate the method.

The authors evaluate research and development in the design of expert systems for the audit domain, providing an overview of the domain of expert judgment involved in the audit process. A framework used to present and analyze work to date and to guide future efforts is constructed. Methods of knowledge acquisition being used to develop audit applications are examined. The authors address knowledge representation for the audit domain. Most systems developed for auditing are rule-based systems. Currently, however, other representational forms are being tested that may be more appropriate for dealing with audit complexity. Progress being made in expert system validation is considered, and the current state of development and directions for future efforts are briefly discussed.>

The integration and coupling that result from networking, database systems, and real-time applications are serving to increase concern about security and control in modem computing environments. The establishment and effective monitoring of controls in such settings is essential to guarantee data integrity and to prevent undesired intrusion. It is proposed that the monitoring of system controls is essentially an information-processing activity and that the relational model is a useful formalization of information gathered from central monitoring. Moreover, we posit that the universal data view facilitates auditor analysis without imposing the requirement that he or she have explicit knowledge of the structure of the relational database of monitoring information.

Analysts from many disciplines are actively involved in the construction of models which will enable better understanding of system structure, improve system design, and aid in planning and decision making. One important obstacle to achieving these ends is the frequent necessity to involve resource people in the modelling process. Such persons are typically expert concerning a particular segment of system operation, but are not as a rule skilled model builders and may have only sketchy knowledge of the overall system structure. Thus, the analyst's challenge is to provide a means of working systematically to establish relationships among fragmented inputs from a body of experts. This paper reports on a computer aid which can be useful in combining the skills of analysts and resource people in the synthesis and design of complex systems. Illustrative applications to information systems and health care planning are included.

Possibly the most revolutionary development to impact upon decisionmaking has been the electronic computer. Its capability to process large volumes of data rapidly, to make complex computations in fractions of seconds, and to provide outputs in a variety of formats has markedly expanded the boundaries of decisionmaking. Notwithstanding such progress, the computer's contribution to decisionmaking has thus far been largely limited to routine and repetitive operations which lend themselves to the development of standard decisionmaking procedures. This has been an important contribution, yet most decisionmaking situations, particularly at higher levels of management, are unique and ill structured requiring flexible and adaptive responses. Man-machine systems are thought to offer potential for assisting management in meeting these requirements. The study reported here examines comparative problem-solving performance of individuals communicating directly with the computer (on-line processing) versus individuals communicating with the computer in an intermittent fashion (batch processing). The study is marked by an effort to take up the challenge of management problems having no set solution procedure. Performance measurements include objective functions, understanding of variable interactions, and comparative problem-solving patterns.

In the research and development environment, the classical decision model, even though modified to include multiple objectives or attributes, fails in the following ways. 1) It uses linear attribute weighting functions. 2) It does not incorporate uncertainty distributions in estimates of total value and attribute attainment. 3) It does not facilitate group decision making. A computer-assisted decision model has been developed to utilize decision analysis information more effectively. A hypothetical analysis of advanced nuclear waste management alternatives illustrates its application. A simulation technique predicts the preferred decision based on estimates of attribute levels, utility functions, and an evaluation of the relative importance of the attributes.