SIDE: A Decision Support System Using a Combination of Swarm Intelligence and Data Envelopment Analysis

Introduction

A business organization’s objective is to make better decisions at all levels of the firm to improve performance. Typically organizations are multi-faceted and complex systems that use uncertain information. Therefore, making quality decisions to improve organizational performance is a daunting task. Organizations use decision support systems that apply different business intelligence techniques such as statistical models, scoring models, neural networks, expert systems, neuro-fuzzy systems, case-based systems, or simply rules that have been developed through experience. Managers need a decision-making approach that is robust, competent, effective, efficient, and integrative to handle the multi-dimensional organizational entities. The decision maker deals with multiple players in an organization such as products, customers, competitors, location, geographic structure, scope, internal organization, and cultural dimension (Porter, 1980). Sound decisions include two important concepts: efficiency (return on invested resources) and effectiveness (reaching predetermined goals). However, quite frequently, the decision maker cannot simultaneously handle data from different sources. Hence, we recommend that managers analyze different aspects of data from multiple sources separately and integrate the results of the analysis. This study proposes the design of a multi-attribute-decision-support-system that combines the analytical power of two different tools: data envelopment analysis (DEA) and particle swarm optimization (PSO), one of the major algorithms using swarm intelligence. DEA evaluates and measures the relative efficiency of decision making units that use multiple inputs and outputs to provide non-objective measures without making any specific assumptions about data. On the other hand PSO’s main strength lies in exploring the entire search space. This study proposes a modeling technique that jointly uses the two techniques to benefit from the two methodologies. A major advantage of the DEA approach is that it clearly identifies the important factors contributing to the success of a decision. In addition, I also propose the use of a particle swarm optimization to assess the global minima or maxima that can aid the decision-maker in making decisions regarding the implications of a decision. One of the important characteristics of population-based search algorithms is their ability to improve the exploration of the search space without falling in the pitfalls of local minima and maxima. The objective of this study is to find a minima solution (ideal loan) nearest to a given loan application using a limited number of iterations. However, not all solutions found using PSO may be true minima so we use the DEA algorithm, a benchmarking technique, to assess the worthiness of the closest optimal solution. The DEA model benchmarks the performance value of the loan application (both original and optimized) against a set of previous loans. Thus, a decision maker can easily analyze and understand any decision using the power of the efficiency frontier algorithm (DEA) and global search algorithm (PSO) to analyze the credit-worthiness of a loan application. The rest of the paper is organized along the following lines. In section II, we provide a review of previous studies on loan evaluation. Section III discusses the model that we use in this study. Section IV discusses the data and methodology used in this study. In section V, we provide an empirical analysis of our results. Section VI summarizes and concludes our study.