Application of Machine Learning for Fraud Detection – A Decision Support System in the Insurance Sector

Finding appropriate decision support systems (DSS) development processes and methodologies is a topic that has kept researchers in the decision support community busy for the past three decades at least. Inspired by Gibson and Nolan’s curve (Gibson & Nolan 1974; Nolan, 1979), it is fair to contend that the field of DSS development is reaching the end of its expansion (or contagion) stage, which is characterized by the proliferation of processes and methodologies in all areas of decision support. Studies on DSS development conducted during the last 15 years (e.g., Arinze, 1991; Saxena, 1992) have identified more than 30 different approaches to the design and construction of decision support methods and systems (Marakas, 2003). Interestingly enough, none of these approaches predominate and the various DSS development processes usually remain very distinct and project-specific. This situation can be interpreted as a sign that the field of DSS development should soon enter in its formalization (or control) stage. Therefore, we propose a unifying perspective of DSS development based on the notion of context. In this article, we argue that the context of the target DSS (whether organizational, technological, or developmental) is not properly considered in the literature on DSS development. Researchers propose processes (e.g., Courbon, Drageof, & Tomasi, 1979; Stabell 1983), methodologies (e.g., Blanning, 1979; Martin, 1982; Saxena, 1991; Sprague & Carlson, 1982), cycles (e.g., Keen & Scott Morton, 1978; Sage, 1991), guidelines (e.g., for end-user computer), and frameworks, but often fail to explicitly describe the context in which the solution can be applied.

Finding appropriate decision support systems (DSS) development processes and methodologies is a topic that has kept researchers in the decision support community busy for the past three decades at least. Inspired by Gibson and Nolan’s curve (Gibson & Nolan 1974; Nolan, 1979), it is fair to contend that the field of DSS development is reaching the end of its expansion (or contagion) stage, which is characterized by the proliferation of processes and methodologies in all areas of decision support. Studies on DSS development conducted during the last 15 years (e.g., Arinze, 1991; Saxena, 1992) have identified more than 30 different approaches to the design and construction of decision support methods and systems (Marakas, 2003). Interestingly enough, none of these approaches predominate and the various DSS development processes usually remain very distinct and project-specific. This situation can be interpreted as a sign that the field of DSS development should soon enter in its formalization (or control) stage. Therefore, we propose a unifying perspective of DSS development based on the notion of context. In this article, we argue that the context of the target DSS (whether organizational, technological, or developmental) is not properly considered in the literature on DSS development. Researchers propose processes (e.g., Courbon, Drageof, & Tomasi, 1979; Stabell 1983), methodologies (e.g., Blanning, 1979; Martin, 1982; Saxena, 1991; Sprague & Carlson, 1982), cycles (e.g., Keen & Scott Morton, 1978; Sage, 1991), guidelines (e.g., for end-user computer), and frameworks, but often fail to explicitly describe the context in which the solution can be applied.

Machine learning is a useful technology for decision support systems and assumes greater importance in research and practice. Whilst much of the work focuses technical implementations and the adaption of machine learning algorithms to application domains, the factors of machine learning design affecting the usefulness of decision support are still understudied. To enhance the understanding of machine learning and its use in decision support systems, we report the results of our content analysis of design-oriented research published between 1994 and 2013 in major Information Systems outlets. The findings suggest that the usefulness of machine learning for supporting decision-makers is dependent on the task, the phase of decision-making, and the applied technologies. We also report about the advantages and limitations of prior research, the applied evaluation methods and implications for future decision support research. Our findings suggest that future decision support research should shed more light on organizational and people-related evaluation criteria.

Data bias is a critical challenge in machine learning applications within the financial and insurance sectors, as it can lead to misleading risk assessments and inaccurate predictive models. A prevalent source of bias in real‐world datasets is the imbalanced distribution of classes, which is particularly problematic in fraud detection, credit risk assessment, and claim prediction. Traditional approaches to handling imbalanced data often rely on undersampling or oversampling techniques. However, these methods may generate unrealistic minority class samples or fail to perform effectively when dealing with extreme class imbalances. In this paper, we propose a configurable technique based on the underbagging method, integrated with a classifier for highly imbalanced datasets. Our approach is designed to enhance the predictive accuracy of the minority class while maintaining robust performance for the majority class. We incorporate our methodology into a classification ensemble framework and evaluate its effectiveness by comparing it against 100 combinations of 10 different oversampling and undersampling techniques applied to 10 different machine learning algorithms. The evaluation is conducted on two highly imbalanced real‐world datasets: one related to auto insurance claims and another focused on credit card fraud detection. Our statistical analysis demonstrates that Balanced Underbagged Ensemble achieves superior classification performance in terms of recall for both classes, regardless of the base machine learning model used within the ensemble. Furthermore, our method finds an optimal balance between classification performance and computational efficiency.

I: General Issues.- Modeling Knowledge: Model-based Decision Support and Soft Computations.- Benefits of Decision Support Using Soft Computing.- Evolving Connectionist-based Decision Support Systems.- An Agent-based Soft Computing Society with Application Applications in Financial Investment Planning.- A Rough Sets/Neoral Networks Approach to Knowledge Discovery for the Development of Decision Support Systems.- II: Applications and Implementations.- Decision Support Systems in Healthcare: Emerging Trends and Success Factors.- A View of Public and Private Sectors for Taiwan's BOT Project Financing Using Fuzzy Multi-Criteria Methods.- Relational Structures for the Analysis of Decision Information in an Electronic Market.- A Fuzzy Evaluation Model: A Case for Intermodal Terminals in Europe.- Application of Kemeny's Median for Group Decision Support.- An Internet-based Group Decision and Consensus Reaching Support System.- Limpio: A DSS for the Urban Waste Collection Problem.- A Decision Support System for Air Quality Control Based on Soft Computing Methods.- Multicriteria Genetic Tuning for the Optimization and Control of HVAC Systems.- Intelligent Information Systems for Crime Analysis.- Application of Fuzzy Decision Trees to Reservoir Recognition.- Introducing SACRA: A Decision Support System for the Construction of Cattle Diets.- Prediction of Parthenium Weed Dispersal Using Fuzzy Logic on GIS Spatial Image.

The expanding quantity of health data and the complexity of its applications are pointing to the need for greater application of computer resources to provide support for decision-making in public health and clinical practice. Decision support and expert systems, as illustrated by the immunization-forecasting program IMM/Serve, offer such support, both now and in the future. Would-be developers of such systems, however, must recognize that the systems are both inherently complex and work-intensive in development. Successful decision support and expert systems require incorporation of comprehensive knowledge and sound logic, extensive testing by use of a variety of methods, and consideration of the nature of the decision-making to be supported and the appropriateness of the environment in which such systems will be placed, including the willingness of users to participate in the development process. Clearly, decision support systems can be appropriate for a number of potential applications in public health practice, including analysis of surveillance data, resource management, and the dissemination of practice guidelines.

This paper critically analyses the nature and state of decision support systems (DSS) research. To provide context for the analysis, a history of DSS is presented which focuses on the evolution of a number of sub-groupings of research and practice: personal DSS, group support systems, negotiation support systems, intelligent DSS, knowledge management-based DSS, executive information systems/business intelligence, and data warehousing. To understand the state of DSS research an empirical investigation of published DSS research is presented. This investigation is based on the detailed analysis of 1,020 DSS articles published in 14 major journals from 1990 to 2003. The analysis found that DSS publication has been falling steadily since its peak in 1994 and the current publication rate is at early 1990s levels. Other findings include that personal DSS and group support systems dominate research activity and data warehousing is the least published type of DSS. The journal DSS is the major publishing outlet; US ‘Other’ journals dominate DSS publishing and there is very low exposure of DSS in European journals. Around two-thirds of DSS research is empirical, a much higher proportion than general IS research. DSS empirical research is overwhelming positivist, and is more dominated by positivism than IS research in general. Design science is a major DSS research category. The decision support focus of the sample shows a well-balanced mix of development, technology, process, and outcome studies. Almost half of DSS papers did not use judgement and decision-making reference research in the design and analysis of their projects and most cited reference works are relatively old. A major omission in DSS scholarship is the poor identification of the clients and users of the various DSS applications that are the focus of investigation. The analysis of the professional or practical contribution of DSS research shows a field that is facing a crisis of relevance. Using the history and empirical study as a foundation, a number of strategies for improving DSS research are suggested.

Web Services has provided the opportunity for organizations to have a platform independent, Decision Support System (DSS) that is always available anywhere at anytime. Although this kind of usage of DSS is relatively new, but it is shown that it would be a valuable technology for e-business. In this paper, we conducted a comprehensive Web Services technology review from business application perspectives and proposed a tentative framework for integrating Web Services with traditional DSS in order to offer online DSS services. Some of the limitations of Web Services particularly in handling transactional tasks have also been addressed.

Data processing is crucial in the insurance industry, due to the important information that is contained in the data. Business Intelligence (BI) allows to better manage the various activities as for companies working in the insurance sector. Business Intelligence based on the Decision Support System (DSS), makes it possible to improve the efficiency of decisions and processes, by improving them to the individual characteristics of the agents. In this direction, Key Performance Indicators (KPIs) are valid tools that help insurance companies to understand the current market and to anticipate future trends. The purpose of the present paper is to discuss a case study, which was developed within the research project "DSS / BI HUMAN RESOURCES", related to the implementation of an intelligent platform for the automated management of agents' activities. The platform includes BI, DSS, and KPIs. Specifically, the platform integrates Data Mining (DM) algorithms for agent scoring, K-means algorithms for customer clustering, and a Long Short-Term Memory (LSTM) artificial neural network for the prediction of agents KPIs. The LSTM model is validated by the Artificial Records (AR) approach, which allows to feed the training dataset in data-poor situations as in many practical cases using Artificial Intelligence (AI) algorithms. Using the LSTM-AR method, an analysis of the performance of the artificial neural network is carried out by changing the number of records in the dataset. More precisely, as the number of records increases, the accuracy increases up to a value equal to 0.9987.

The purpose of this chapter is to discuss how characteristics of a decision support system (DSS) interact with characteristics of a task to affect DSS use and decision performance. This discussion is based on the motivational framework developed by Chan (2005) and the studies conducted by Chan (2009) and Chan et al. (2009). The key constructs in the motivational framework include task motivation, user perception of DSS, motivation to use a DSS, DSS use, and decision performance. This framework highlights the significant role of the motivation factor, an important psychological construct, in explaining DSS use and decision performance. While DSS use is an event where users place a high value on decision performance, the Technology Acceptance Model (TAM) and the Unified Theory of Acceptance and Use of Technology (UTAUT) do not explicitly establish a connection between system use and decision performance. Thus, Chan (2005) includes decision performance as a construct in the motivational framework rather than rely on the assumption that DSS use will necessarily result in positive outcomes (Lucas & Spitler, 1999; Venkatesh et al., 2003). This is an important facet of the framework because the ultimate purpose of DSS use is enhanced decision performance. Chan (2009) tests some of the constructs in the motivational framework. Specifically, the author examines how task motivation interacts with DSS effectiveness and efficiency to affect DSS use. As predicted, the findings indicate that individuals using a more effective DSS to work on a high motivation task increase usage of the DSS, while DSS use does not differ between individuals using either a more or less effective DSS to complete a low motivation task. The results also show significant differences for individuals using either a more or less efficient DSS to complete a low motivation task, but no significant differences between individuals using either a more or less efficient DSS to perform a high motivation task only when the extent of DSS use is measured dichotomously (i.e., use versus non-use). These findings suggest the importance of task motivation and corroborate the findings of prior research in the context of objective (i.e., computer recorded) rather than subjective (self-reported) DSS use. A contribution of Chan’s (2009) study is use of a rich measure of DSS use based on Burton-Jones and Straub’s (2006) definition of DSS use as an activity that includes a user, a DSS, and a task. Chan et al. (2009) extends the motivational framework by investigating the alternative paths among the constructs proposed in the framework. Specifically, the authors test the direct

Geographical Information Systems (GIS) are gaining increasing importance and widespread acceptance as tools for decision support in land, infrastructure, resources, environmental management and spatial analysis, and in urban and regional development planning. GIS assist in the preparation, analysis, display, and management of geographical data. It is in the analysis and display functions that GIS meet Decision Support Systems (DSS). DSS analyse and support decisions through the formal analysis of alternative options, their attributes vis-a-vis evaluation criteria, goals or objectives, and constraints. DSS functions range from information retrieval and display, filtering and pattern recognition, extrapolation, inference and logical comparison, to complex modelling. The use of model-based information and DSS, and in particular of interactive simulation and optimization models that combine traditional modelling approaches with new expert systems techniques of Artificial Intelligence (AI), dynamic computer graphics and geographical information systems, is demonstrated in this chapter with application examples from technological risk assessment, environmental impact analysis, and regional development planning. With the emphasis on an easy-to-understand visual problem representation, using largely symbolic interaction and dynamic images that support understanding and insight, these systems are designed to provide a rich and directly accessible information basis for decision support and planning.

Robert W. Blanning is Professor of Management Information Systems at the Owen Graduate School of Management at Vanderbilt University. He holds a B.S. in Physics from the Pennsylvania State University, an M.S. in Operations Research from the Case Institute of Technology, and a Ph.D. from the University of Pennsylvania, specializing in operations research and management information systems. He has been a member of the faculties of the School of Business at New York University and the Wharton School at the University of Pennsylvania. His teaching and research interests are in model management systems, information economics, and the management applications of artificial intelligence. He has published in Management Science, Decision Sciences, Communications of the ACM, Naval Research Logistics Quarterly, Decision Support Systems, Information and Management, Omega, Policy Analysis and Information Systems, International Journal of Policy and Information, Human Systems Management, Journal of Information Science, Long Range Planning, and Technological Forecasting and Social Change. He has presented papers at the National Computer Conference, the International Conference on Decision Support Systems, the Hawaii International Conference on Systems Sciences, the International Workshop on Expert Database Systems , and the International Workshop on Artificial Intelligence in Economics and Management. He is a member of the Board of Editors of JMIS, a member of the Editorial Boards of Decision Support Systems and Information Systems Research, and Associate Editor of Information and Decision Technologies and in the Decision Support Systems Department of Management Science.

Remote decision support system: a distributed information management system

This study addresses the impact of Group Decision Support Systems (GDSS) on expert system development by multiple Domain Experts. Current approaches to building expert systems rely heavily on knowledge acquisition and prototyping by a Knowledge Engineer working directly with the Domain Expert. Although the complexity of knowledge domains and new organizational approaches demand the involvement of multiple experts, standard procedures limit the ability of the Knowledge Engineer to work with more than one expert at a time. Group Decision Support Systems offer a networked computerized environment for group work activities, in which multiple experts may express their ideas concurrently and anonymously through the electronic channel. GDSS systems have been widely used in other applications to support idea generation, conflict management, and the organizing, prioritizing, and synthesizing of ideas. The effects of many group process and technical factors on GDSS have been widely studied and documented. A review of the literature on expert systems, GDSS, and GDSS in relation to expert systems was conducted. Knowledge gained from this review was applied in the construction of an exploratory research model intended to provide the necessary breadth to identify factors worthy of future, more statistically-based, investigation. Domain Experts represented by college students were charged with developing and prioritizing ideas for creating a pre-prototypical expert system. The treatment group worked in a GDSS environment with a facilitator; a control group worked with a facilitator but without the assistance of GDSS. Each group then exchanged facilitators and technology to address another real-life problem. Additional groups worked with GDSS over time, addressing both problems. Data were gathered, analyzed and discussed relating to group efficiency factors, group process factors, attitudinal factors, and product quality factors. Independent Knowledge Engineers and Domain Experts evaluated the validity and verifiability of the group products. Analysis focused on the effect of GDSS in facilitating the acquisition and structuring of ideas for expert systems by multiple Domain Experts.

The real estate industry frequently faces challenges in achieving accurate property valuations, a crucial factor in property transactions. Evaluating commercial real estate tenants is crucial for various stakeholders in the real estate industry, such as property owners, investors, lenders, and property managers. Due to the challenges, this study aims to present a novel Decision Support (DS) software developed to address those challenges by integrating both front-end and back-end systems for seamless, data-driven property assessments. DSS can play a critical role in ensuring accurate property valuations – a key aspect of real estate transactions that often relies on outdated or static data. While the front-end side offers an intuitive interface for users to input property data and receive near-instant results, the back-end side is powered by advanced machine learning algorithms to process this information efficiently. This dual system can ensure that users not only receive accurate valuations but also benefit from a seamless and intuitive user experience. In this paper, we outline the development process, the technologies which we have used, and the evaluation methods which was applied to ensure that the DSS is reliable and effective. The Decision Support System’s software for property valorization was developed through comprehensive and multi-stage processes. This system not only enhances the accuracy of valuations but also streamlines the overall decision-making process. The DSS addresses several key challenges in property valuation, including fluctuating market conditions and the complexities of data interpretation. Its scalable design and the continuous learning capabilities of its AI models ensure that the system remains relevant and accurate as market dynamics change. The development of the Decision Support Software (DSS) for property valorization represents a significant step forward in real estate technology. The software’s ability to provide accurate, real-time valuations, combined with its user-friendly interface and detailed market insights, positions it as a valuable tool for real estate professionals, investors, and regulators alike.