This is a collection of slides presented at IFORS 2014, sessions FA-42 and FB-42, on July 18, 2014. All documents are reproduced by authors' permissions.
Abstract: DEX is a qualitative multi-attribute decision method, aimed at evaluation and analysis of decision alternatives. Conceptually, DEX is a combination of multi-criteria decision analysis and expert systems. DEX’s models are hierarchical and composed of qualitative (symbolic) variables, whose interrelations are modeled with decision rules. DEX and the supporting software DEXi were used in many practical applications. The purpose of this work is threefold: (1) formal description of DEX, (2) overview of its practical applications, and (3) research and development challenges for the future.
Abstract: Evaluation of patient’s health status is usually based on different signs and symptoms which have to be aggregated in final estimate of patient’s status. This can be viewed as a multi attribute decision making (MADM) problem. In this contribution we present the implementation of Henderson’s theoretical model of basic living activities for patient health status evaluation. Qualitative MADM methodology DEX is used, which facilitates user friendly acquisition and explanation of expert knowledge. Practical evaluation of our solution confirmed the added value in transparency of patient’s status.
Abstract: The topic of the paper is comparative evaluation of various energy options. It is treated from the strategic evaluation point of view focusing on the constituents of sustainability appraisal. The latter examines differences between the approach, which builds on specific indicators like climate change, ecology, air quality, health and well being, etc., and the approach, which rather applies more general interpretation based on rationality, feasibility, and uncertainties of energy options. The evaluation was supported by qualitative multi-attribute modeling method DEX.
Abstract: We provide a solution for ranking of qualitative multi-attribute options modeled with DEX methodology. In DEX, the attributes form a hierarchical structure which solves the problem of sorting the options into preferentially ordered classes. To obtain full option ranking within each of the classes for nonlinear non-monotone options, we employ copula functions. The property of full option ranking is relaxed in cases of symmetric attributes, and its solved by introducing exchangeable copulas. The solution is demonstrated on real and artificial cases.
Abstract: DEX is a qualitative decision support method aimed at evaluation and analysis of decision alternatives. Many real decision problems are based on relational properties between at least two types of entities, and require a combination of numeric and symbolic attributes. For example, evaluation of bank’s reputational risk has relations between banks, bank’s counterparts and clients, and financial products. In this work we address the task of extending DEX to facilitate evaluation of relationally connected decision alternatives, described with a combination of numeric and symbolic attributes.
Abstract: High Performance Computing offered as a cloud service is regarded as one of the key competitiveness boosters for SMEs, particularly manufacturing. However, business models are not yet explored, and technology adoptance is in its early stages. In order to explore and support new business ideas there is need for assessment of the SMEs readiness and market viability. Based on theory and practice we are proposing a qualitative multi-attribute model for SMEs’ cloud HPC adoption assessment. The model will be verified on a set of experiments conducted within several EU projects in I4MS initiative.
Abstract: Business intelligence (BI) systems offer wide variety of functionalities for data retrieval, analysis and visualization, improving the quality of business decision-making. Still, the usage analyses reveal rather low BI adoption. In this contribution, we discuss the complexity of BI systems and approaches to make them user- and context-specific. We analyse results of a case study for determining different types of BI users from system-user interaction traces using clustering, and adaptation of data cubes to user needs by using multiple criteria decision analysis.
Abstract: Multiple-attribute selection decisions require gathering information using observations that can provide data about only one of the attributes at a time. When resources are scarce, the decision-maker must choose which attributes to observe (sample) in a way that maximizes the likelihood that the correct alternative will be selected. This talk will describe sample allocation policies for multiple-attribute selection decisions with attribute values estimated through pass-fail testing (Bernoulli trials) and those estimated by techniques that are subject to normally distributed measurement error.