Addressing Interdependence and Data Inadequacy in Multi-Criteria Decision Making with an Application to Singapore Casino Industry

Cai Zhilin

Abstract

Interdependences among different criteria for decision making is common in many complex real-life scenarios. Data inadequacy can be a result of poor research method, the inherent complexity of the context and budget constraint. This is a difficulty faced by scholars in studying real-life problems. Neglecting these problems may give invalid result since the model may not be representative and data input may not be accurate. The paper proposes a seven-step decision support process for decision making in complex problems with potential criteria interdependences and data inadequacy.

The proposed decision support process starts from identifying appropriate alternatives for the subject of interest. Then, relevant criteria affecting the goal will be identified under the specific context being studied. The problem will be mapped into a hierarchical structure adopting Analytic Hierarchy Process (AHP) technique. Subsequently, decision maker would decide whether the interactions among criteria are significant and whether there is a need to reflect these interactions in the model. If so, the model may be further modified so that the criteria can be organised in a network with the use of Analytic Network Process (ANP). Next, the decision maker would review the process and examine the severity of data inadequacy, which may lead to the need of introducing fuzziness in the model. The model is then modified through the use of fuzzy ANP and the decision support process ends. A case application to Singapore casino industry is done to illustrate how the proposed decision support process works under this scenario. AHP is first adopted to represent the situation. The final recommendation given by AHP is not distinct. A further step is then made by considering inter-dependencies among different criteria and re-map the system in a network with the use of ANP technique. The result has shown that ANP is able to produce a more discriminative result in giving the final recommendation. This may assist the decision maker in the decision making process. Afterwards, fuzzy ANP technique is used to study the same problem so that possible deviation in values caused by data inadequacy is taken into account. This allows decision makers to reflect their confidence level in the data input by adjusting the fuzzy membership functions.

After performing these three techniques, suggestions are made for a possible model extension to include fuzzy logic in a compressed ANP structure containing only the goal, criteria and alternative levels. This can provide an adequate recommendation without going through tedious computations for a full-scale network problem.

The proposed decision support is flexible and it takes into account interdependence among criteria, the problem of data inadequacy and the confidence level on data input. The case application has shown its practicality in solving real-life problems. With the use of the proposed decision support process, the study on casino problem is flexible can cover the use of intangible criteria, possible estimation error and the confidence level on judgement. This is an improvement from the past studies on casino.