Consideration of the social context of auditors’ reliance on expert system output during evaluation of loan loss reserves

Auditors are accountable for judgments made within the social context of the accounting firm. Tetlock (1985) states that decision makers often use the acceptability heuristic to cope with accountability. According to this heuristic, individuals make decisions which they are reasonably confident will be acceptable by others to whom they are accountable. When auditors form judgments with the aid of expert system output, they must determine the appropriate level of reliance on the expert system output. Since the expert system output is based on the input of experts, auditors may decide the output is ‘acceptable‘ and overrely on the output. In addition, because of the conservative nature of the accounting firm, expert system output which is negative may be viewed as more acceptable than positive output leading to greater overreliance. The results indicate that auditors do overrely on expert system output and rely to a greater degree on output which is negative versus output which is positive. Copyright © 1999 John Wiley & Sons, Ltd.     

Hybrid knowledge integration using the fuzzy genetic algorithm: prediction of the Korea stock price index

This paper proposes the hybrid knowledge integration mechanism using the fuzzy genetic algorithm for the optimized integration of knowledge from several sources such as machine knowledge, expert knowledge and user knowledge. This mechanism is applied to the prediction of the Korea stock price index. Machine knowledge is generated by applying neural networks to technical indicators, while expert knowledge and user knowledge are generated from the evaluations of external factors that affect the stock market. Cooperative knowledge is generated from the weighted sum of these sources using a genetic algorithm. Experimental results show that the hybrid mechanism can provide more accurate and less ambiguous results. It means that this mechanism is useful in integrating knowledge from multiple sources for an unstructured environment such as the stock market. Copyright © 2004 John Wiley & Sons, Ltd.     

Combining Expert Systems and Hypertext: Opportunities and Obstacles

Hypertext discussions are occurring more frequently at expert systems conferences. Hypertext is not an expert system language or expert system shell, but instead it is simultaneously a method of storing and retrieving data. The growing interest in hypertext in the expert system domain is because the combination of hypertext capabilities and expert systems creates a compelling synergistic relationship (Oren, 1987). During a panel discussion at the Second International Symposium on Expert Systems in Business, Finance and Accounting, Bill Swartout compared traditional expert systems to hypertext by saying that the knowledge base (e.g. the production rules) of an expert system can be thought of as formal knowledge and hypertext can be thought of as informal knowledge. This article will demonstrate how this informal knowledge can add power to traditional expert systems by increasing the flexibility of the explanation facilities and thereby the appeal of the system to a broader range of users. The article also will discuss how expert systems can improve the functionality of hypertext systems by adding intelligence to traditional hypertext systems. This article is divided into three sections. The first describes basic hypertext concepts. The second section shows how hypertext can enhance expert systems and, conversely, how expert systems can enhance hypertext systems. The final section discusses some potential problems and concerns that must be considered when designing a hypertext system.     

A Case Study and Field Evaluation of EDP-XPERT

This paper reports on the continued evaluation of EDP-XPERT, and expert system for assisting Computer Audit Specialists (CASs) in evaluating the reliability of EDP controls in advanced computer environments. The current evaluation of the system involved four phases. First, Senior CASs used the system on two case studies. Second, it was then used by the Senior CASs to evaluate recently completed audits. Third, a sensitivity analysis of the system's output was conducted. Fourth, the knowledge base was reprogrammed into another expert system shell and the Senior CASs' responses on the two cases and the audits were re-run. The results indicated that EDP-XPERT performed well on the case studies but that performance declined on the audits. The sensitivity analysis showed that EDP-XPERT's conclusions were affected by moderate perturbations of the CASs' input judgments. Finally, reprogramming the knowledge base into a more flexible expert system shell substantially improved the performance of the system's judgements.     

Automated trading with boosting and expert weighting

We propose a multi-stock automated trading system that relies on a layered structure consisting of a machine learning algorithm, an online learning utility, and a risk management overlay. Alternating decision tree (ADT), which is implemented with Logitboost, was chosen as the underlying algorithm. One of the strengths of our approach is that the algorithm is able to select the best combination of rules derived from well-known technical analysis indicators and is also able to select the best parameters of the technical indicators. Additionally, the online learning layer combines the output of several ADTs and suggests a short or long position. Finally, the risk management layer can validate the trading signal when it exceeds a specified non-zero threshold and limit the application of our trading strategy when it is not profitable. We test the expert weighting algorithm with data of 100 randomly selected companies of the S&P 500 index during the period 2003–2005. We find that this algorithm generates abnormal returns during the test period. Our experiments show that the boosting approach is able to improve the predictive capacity when indicators are combined and aggregated as a single predictor. Even more, the combination of indicators of different stocks demonstrated to be adequate in order to reduce the use of computational resources, and still maintain an adequate predictive capacity.     

An adaptive hierarchical fuzzy logic system for modelling of financial systems

In this paper an intelligent hierarchical fuzzy logic system using genetic algorithms for the prediction and modelling of interest rates in Australia is developed. The proposed system uses a hierarchical fuzzy logic system in which a genetic algorithm is used as a training method for learning the fuzzy rules knowledge bases that are used for prediction of interest rates in Australia. A hierarchical fuzzy logic system is developed to model and predict three‐month (quarterly) interest rate ?uctuations. The system is further trained to model and predict interest rates for six‐month and one‐year periods. The proposed system is developed with ?rst two, three, then four and ?nally ?ve hierarchical knowledge bases to model and predict interest rates. A novel architecture called a feed forward fuzzy logic system using fuzzy logic and genetic algorithms is also developed to predict interest rates. A back‐propagation hierarchical neural network system is also developed to predict interest rates for three‐month, six‐month and one‐year periods. The results obtained from these two systems are then compared with the hierarchical fuzzy logic system results and conclusions are drown on the accuracy of all systems for prediction of interest rates in Australia. Copyright © 2004 John Wiley & Sons, Ltd.     

Refining the behavior of multiple expert systems: a concept and empirical results

Expert System Refinement (ESR) is introduced as a means to automatically refine the performance of one or more expert systems. The ESR concept is based on Holland’s learning classifier systems and a method for integrating multiple expert systems. Through users’ feedback about the usefulness/correctness of the integrated expert system’s recommendations, ESR enables behaviors of both individual expert systems as well as the integrated system to improve over time. The ESR concept is tested on a German Credit Database. This empirical evidence suggests that the ESR concept can be usefully applied in automating the process of expert system refinement and multiple expert systems integration. © 1998 John Wiley & Sons, Ltd.     

Changing Impact of Shocks: A Time-Varying Proxy SVAR Approach

In this paper, we extend the Bayesian Proxy vector autoregression (VAR) model to incorporate time variation in the parameters. A novel Metropolis-within-Gibbs sampling algorithm is provided to approximate the posterior distributions of the model's parameters. Using the proposed algorithm, we estimate the time-varying effects of taxation shocks in the United States and the United Kingdom and find evidence for a decline in the impact of these shocks on output growth.     

A STOCHASTIC SETTING TO BANK FINANCIAL PERFORMANCE FOR REFINING EFFICIENCY ESTIMATES

This study contributes to develop a framework to measure the financial performance of banks in a stochastic setting. The framework comprises several steps, the first of which is the development of a financial performance measurement model to evaluate a bank's financial performance using a set of factors from the CAMEL (Capital adequacy, Assets, Management Capability, Earning and Liquidity) system. Second, the stochastic setting of the efficiency measurement is handled using the data collection budget allocation approach, whereby Monte Carlo simulations are used to analyse additional generated data and a genetic algorithm is used to refine the accuracy of the efficiency estimates. The results show that the accuracy of the model is greatly improved using the proposed approach. In contrast to the conventional deterministic model, the proposed framework is more useful to managers in determining the bank's future financial operations to improve the overall financial soundness of the bank. Copyright © 2014 John Wiley & Sons, Ltd.     

Integrating genetic algorithms and spreadsheets: a capital budgeting application

The role of the tax system in generating interactions between the post‐tax cash flows of different projects is discussed. When such interactions can occur, the capital budgeting process should be based around project combinations rather than individual projects. Evaluation of a project combination in net present value terms can easily be done using a spreadsheet. If the number of individual projects is large, then project combinations can be generated and an optimum combination of projects searched for using a genetic algorithm. The genetic algorithm approach has an advantage over alternative computational approaches, such as mixed integer programming, because of the more understandable representation of the problem it allows. Copyright © 2007 John Wiley & Sons, Ltd.     

Experiences with EXGAME: an expert system for playing a competitive business game

This paper looks at expert systems in management, by using a business game as an experimental vehicle. An expert system called EXGAME was developed to play a business game, which is normally played by students, with minimal human intervention. This paper concentrates on the effectiveness of EXGAME as compared with human players for tasks at different levels. EXGAME was able to replace human players in decision making at the operational level, and indeed outperform them. However, it proved to be impractical to replace human input at the strategic level. The paper also sheds some light on the problems of trying to build an expert system when there is no real expert. A combination of a modular knowledge-base structure and a process of ‘learning by experimentation’ was effective in this case; it is suggested that this may be an appropriate development strategy in other similar situations. © 1998 John Wiley & Sons, Ltd.     

An Analysis of Potential Legal Liability Incurred Through Audit Expert Systems

The introduction of expert systems technology into the audit environment has opened a new avenue of auditor legal liability. This paper examines the potential impact expert systems will have on auditor liability. The presentation of this new avenue of auditors' legal liability explores both the potential for litigation under failure of auditor/expert system collaboration to yield prudent decisions and the failure to use an available expert system. The risks evolving from failure to use an available expert system include the possibility that the system could be used against the auditor in the courtroom. While case law will ultimately determine the bounds of this liability, this paper acquaints the reader with the important legal issues involved and the varied outcomes that could emerge. It should also be noted that while the specific example presented in this paper relates to the audit profession, the legal concepts are of equivalent concern to other professions enduring broad implementation of expert systems.     

A Comparison of State-of-the-Art Classification Techniques for Expert Automobile Insurance Claim Fraud Detection

Several state‐of‐the‐art binary classification techniques are experimentally evaluated in the context of expert automobile insurance claim fraud detection. The predictive power of logistic regression, C4.5 decision tree, k‐nearest neighbor, Bayesian learning multilayer perceptron neural network, least‐squares support vector machine, naive Bayes, and tree‐augmented naive Bayes classification is contrasted. For most of these algorithm types, we report on several operationalizations using alternative hyperparameter or design choices. We compare these in terms of mean percentage correctly classified (PCC) and mean area under the receiver operating characteristic (AUROC) curve using a stratified, blocked, ten‐fold cross‐validation experiment. We also contrast algorithm type performance visually by means of the convex hull of the receiver operating characteristic (ROC) curves associated with the alternative operationalizations per algorithm type. The study is based on a data set of 1,399 personal injury protection claims from 1993 accidents collected by the Automobile Insurers Bureau of Massachusetts. To stay as close to real‐life operating conditions as possible, we consider only predictors that are known relatively early in the life of a claim. Furthermore, based on the qualification of each available claim by both a verbal expert assessment of suspicion of fraud and a ten‐point‐scale expert suspicion score, we can compare classification for different target/class encoding schemes. Finally, we also investigate the added value of systematically collecting nonflag predictors for suspicion of fraud modeling purposes. From the observed results, we may state that: (1) independent of the target encoding scheme and the algorithm type, the inclusion of nonflag predictors allows us to significantly boost predictive performance; (2) for all the evaluated scenarios, the performance difference in terms of mean PCC and mean AUROC between many algorithm type operationalizations turns out to be rather small; visual comparison of the algorithm type ROC curve convex hulls also shows limited difference in performance over the range of operating conditions; (3) relatively simple and efficient techniques such as linear logistic regression and linear kernel least‐squares support vector machine classification show excellent overall predictive capabilities, and (smoothed) naive Bayes also performs well; and (4) the C4.5 decision tree operationalization results are rather disappointing; none of the tree operationalizations are capable of attaining mean AUROC performance in line with the best. Visual inspection of the evaluated scenarios reveals that the C4.5 algorithm type ROC curve convex hull is often dominated in large part by most of the other algorithm type hulls.     

Linear models for minimizing misclassification costs in bankruptcy prediction

This paper illustrates how a misclassification cost matrix can be incorporated into an evolutionary classification system for bankruptcy prediction. Most classification systems for predicting bankruptcy have attempted to minimize misclassifications. The minimizing misclassification approach assumes that Type I and Type II error costs for misclassifications are equal. There is evidence that these costs are not equal and incorporating costs into the classification systems can lead to better and more desirable results. In this paper, we use the principles of evolution to develop and test a genetic algorithm (GA) based approach that incorporates the asymmetric Type I and Type II error costs. Using simulated and real-life bankruptcy data, we compare the results of our proposed approach with three linear approaches: statistical linear discriminant analysis (LDA), a goal programming approach, and a GA-based classification approach that does not incorporate the asymmetric misclassification costs. Our results indicate that the proposed approach, incorporating Type I and Type II error costs, results in lower misclassification costs when compared to LDA and GA approaches that do not incorporate misclassification costs. Copyright © 2001 John Wiley & Sons, Ltd.     

The development of an expert system for managerial evaluation of internal controls

Both practitioners and researchers have devoted signi?cant effort to the study of decision aids, especially expert systems, to assist auditors in internal control evaluations. In addition to being used as a decision aid, researchers have long contended that expert systems could be used to train non‐expert users. Even though the professional accounting literature makes it clear that responsibility for maintaining an effective internal control system rests with management rather than auditors, the focus to date has been on expert systems aimed at assisting/training auditors, not an organization's management. In contrast, this study focuses on management as users of an expert system for internal control evaluation. We describe the development process, explain how the resulting system was evaluated, and discuss results of that evaluation. These results suggest that such a system gives a new way to help managers increase effectiveness and ef?ciency of a critical organizational process: the evaluation of internal controls. Copyright © 2004 John Wiley & Sons, Ltd.     

A validated expert system for decision making in corporate recovery

This paper describes INSOLVE—an expert system for corporate recovery decisions. INSOLVE was built to understand the decision-making processes of corporate recovery experts who deal with companies in financial difficulties. INSOLVE has been developed using a multi-phase process similar to that widely adopted in software engineering. The expert system is described in terms of the assessment task and interpretation models of CommonKADS. The detailed results of the validation of INSOLVE with 17 experts show that it is an accurate model of human expertise in this domain. Copyright © 1999 John Wiley & Sons, Ltd.     

A fuzzy logic-driven multiple knowledge integration framework for improving the performance of expert systems

To maintain a high performance in an ill-structured situation, expert systems should depend on multiple sources of knowledge rather than a single type. For this reason, we propose multiple knowledge integration by using a fuzzy logic-driven framework. Types of knowledge being considered here are threefold: machine, expert and user. Machine knowledge is obtained by a back- propagation neural network model from historical instances of a target problem domain. Expert knowledge is related to interpreting the trends of external factors that seem to affect the target problem domain. User knowledge represents a user’s personal views about information given by both expert knowledge and machine knowledge. The target problem domain of this paper is one-week-ahead stock market stage prediction: Bull, Edged-up, Edged-down, and Bear. Extensive experiments with real data proved that the proposed fuzzy logic-driven framework for multiple knowledge integration can contribute significantly to improving the performance of expert systems. Copyright © 1998 John Wiley & Sons, Ltd.     

Expert Systems for Evaluating Business Opportunities: Implementing the Management Advisor at Krypton Chemical

This paper investigates issues arising in procuring and implementing a commercial high-end expert system, Palladian's Management Advisor. This system is intended for improving corporate decisions about financial investments in new business opportunities, e.g. projects involving new plants, equipment, products, or marketing expenditures. We look at Krypton Chemical's experiences with the Management Advisor from both the customer's and the vendor's viewpoints. The discussion raises some difficult questions about what it takes to make high-end financial expert systems successful. The paper then concludes with recommendations intended to improve design, marketing, procurement and implementation practices with respect to such systems.     

AMTexpert: An Expert Decision Support System for the Corporate Alternative Minimum Tax

AMTexpert is a prototype of an expert decision support system (EDSS) for the corporate alternative minimum tax (AMD. The corporate AMT was introduced as part of the Tax Reform Act of 1986 in order to force US corporations to pay more tax. The complicated and interrelated nature of the AMT system has made compliance and planning even more difficult for corporate taxpayers. Given the complexities of the corporate AMT, significant benefits may accrue from the development of an EDSS. In effect, an EDSS combines the judgmental benefits of an expert system (ES) with the computational benefits of a decision support system (DSS). AMTexpert is menu driven with six interdependent phases. The heart of the system is a ‘scoreboard’ used to compare and evaluate alternative tax positions. Context-sensitive advice and explanations are provided to guide the user. The system was validated and evaluated by two groups of outside experts. AMTexpert demonstrates that an EDSS may be developed from existing tax and systems software.