Integrating Fuzzy C-Means and TOPSIS for performance evaluation: An application and comparative analysis
Introduction
The field of discrete alternative multiple criteria decision analysis (MCDA) and choice have been a mainstay of modeling for a number of years. MCDA has had numerous practical applications in addition to an extensive theoretical history. Managers and decision makers within organizations and people in their everyday lives face a broad variety of decisions that require the consideration and tradeoffs associated with multiple attributes, criteria, and factors. Areas in environmental, economic, operational, strategic, marketing, engineering, design, educational, and psychological disciplines and others have come to rely on some of the latest developments in MCDA. Our research seeks to advance the understanding and development of models in MCDA to make decision making more effective, efficient, and reliable. Using clustering approaches for MCDA is one way of accomplishing this task. We seek to introduce a unique multi-stage approach to address this issue.
In this paper, we introduce the integration of Fuzzy C-Means (FCM), and Technique for Order of Preference by Similarity to Ideal Solution (TOPSIS) for discrete alternative multiple criteria ranking. We have combined the strong points of these two methodologies to construct an integrated approach to rank data objects based on multiple criteria. The combined method minimizes computational effort needed to arrive at rankings since it develops them in two stages (achieving an ‘efficiency’ target for a good MCDA model). First, the FCM algorithm is used categorize similar data objects into clusters. Then TOPSIS is utilized to develop rankings for the clusters and also separate rankings for the data objects within the clusters. To examine the validity of this tool, a comparative analysis will be completed to evaluate the integrated FCM/TOPSIS approach against a novel methodology based on Bayesian inference and a latent class clustering model (Mistry, Sarkis, & Dhavale, 2014). This research aspect will determine the effectiveness and reliability of the proposed FCM/TOPSIS technique. This latent class technique uses Markov Chain Monte Carlo simulation to extract information about the objects that is already present in the multi-characteristic data, but is not readily apparent. As part of the research we found some divergences in the solutions when comparing both approaches using real world data on electronic commerce organizations.
A secondary research goal and contribution of this paper is an evaluation of applying the proposed tool to identify characteristics of company resilience and viability. Specifically, in this paper, we apply FCM/TOPSIS and latent class model to real-world archival data to determine the viability of electronic commerce organizations. The perspectives defined in the well-known Balanced Scorecard method are used to develop the multi-criteria objectives. A unique evaluation technique called a displacement index is used in the comparative analysis and validation/reliability of the joint FCM/TOPSIS methodology. In this research we examine which dimensions of the Balanced Scorecard method would be better predictors of resilience.
This paper contributes to the literature first by introducing the integration of FCM/TOPSIS and secondly, evaluating this approach using real world information and comparing it against a benchmark provided by a robust statistical technique based on Bayesian inference. We also provide a variety of analyses that allow us to evaluate how well various factors or perspectives may perform in this environment. Additionally, potential applications and future research avenues are identified to help build upon this novel approach.
To help accomplish the goals of this paper, we begin by providing a brief background on the use of clustering approaches for MCDA, along with background on FCM/TOPSIS techniques. Then data for an illustrative application of the approach is presented. The section following contains comparative analysis and discussion. The paper concludes with a summary of findings, limitations and direction for future research.
Section snippets
Background
Many approaches exist for MCDA (e.g. see Guitouni & Martel, 1998) and can be categorized on a variety of dimensions including explicit versus implicit alternative valuation, compensatory versus non-compensatory, discrete versus continuous alternative, interactive versus non-interactive, objective versus subjective weighting (Bai and Sarkis, 2012, Wang et al., 2009, Yoon and Hwang, 1995). Applications of MCDA approaches have been completed as stand-alone tools, but many times multi-stage
Illustrative application
We now provide an illustrative application of the Fuzzy C-Means and TOPSIS methodologies to rank performance of actual companies based on multiple criteria. Later in the paper we compare this approach to another approach, a latent class model within a Bayesian framework, to rank the companies using the same set of criteria. The latent class model is selected due to its similarity of utilizing a clustering approach for ranking purposes. We compare the ranking under these two methodologies under
Prediction of future firm performance using FCM/TOPSIS
In this section, we attempt to answer the question of how the FCM/TOPSIS technique can be used to evaluate various performance measure perspectives capabilities to foretell organizational resilience.
Table 8 documents the subsequent history of each company in the two-year period following 1999. Each company’s history was tracked by obtaining relevant announcements and financial information from publicly available databases. As shown in Table 8, eight out of the eleven companies in the
Summary and research contribution
In this paper, we introduce a FCM/TOPSIS approach for ranking objects based on simultaneous consideration of multiple criteria and compare against a new methodology, latent class clustering. Both techniques utilize cluster-based ranking systems and thus a more direct comparison could be made by this methodological similarity. This research helps to expand the scope of utilizing cluster based ranking approaches for multiple criteria decision analysis (MCDA). The specific application focus of
Acknowledgments
This work was partially supported by the National Natural Science Foundation of China Project (71102090); Liaoning Education Department Foundation of China (W2011125).
References (38)
An efficient technique for superfamily classification of amino acid sequences: Feature extraction, fuzzy clustering and prototype selection
Fuzzy Sets and Systems
(2005)- et al.
Enhanced fuzzy clustering algorithm and cluster validity index for human perception
Expert Systems with Applications
(2013) - et al.
Validation criteria for enhanced fuzzy clustering
Pattern Recognition Letters
(2008) - et al.
Combining grey relation and TOPSIS concepts for selecting an expatriate host country
Mathematical and Computer Modelling
(2004) - et al.
Fuzzy C-Means clustering with spatial information for image segmentation
Computerized Medical Imaging and Graphics
(2006) - et al.
Tentative guidelines to help choosing an appropriate MCDA method
European Journal of Operational Research
(1998) - et al.
Fuzzy C-Means and fuzzy swarm for fuzzy clustering problem
Expert Systems with Applications
(2011) - et al.
Cluster center initialization algorithm for K-Means clustering
Pattern Recognition Letters
(2004) - et al.
Effective FCM noise clustering algorithms in medical images
Computers in Biology and Medicine
(2013) - et al.
Fuzzy TOPSIS for group decision making: A case study for accidents with oil spill in the sea
Expert Systems with Applications
(2011)
Multi-criteria analysis using latent class cluster ranking: An investigation into corporate resiliency
International Journal of Production Economics
Compromise solution by MCDM methods: A comparative analysis of VIKOR and TOPSIS
European Journal of Operational Research
Color image segmentation using histogram thresholding – Fuzzy C-Means hybrid approach
Pattern Recognition
Review on multi-criteria decision analysis aid in sustainable energy decision-making
Renewable and Sustainable Energy Reviews
Application of TOPSIS in evaluating initial training aircraft under a fuzzy environment
Expert Systems with Applications
Information and control
Fuzzy Sets
Introduction to machine learning
Supply-chain performance–measurement system management using neighbourhood rough sets
International Journal of Production Research
Green information technology strategic justification and evaluation
Information Systems Frontiers
Cited by (81)
OS-FCM: A semi-supervised clustering approach to investigating consumers’ usage patterns of contactless shopping-delivery (S-D) channel
2024, Expert Systems with ApplicationsSafety in lithium-ion battery circularity activities: A framework and evaluation methodology
2023, Resources, Conservation and RecyclingTemporal dependence and bank efficiency drivers in OECD: A stochastic DEA-ratio approach based on generalized auto-regressive moving averages
2023, Expert Systems with ApplicationsA modified CRITIC with a reference point based on fuzzy logic and hamming distance
2022, Knowledge-Based SystemsAssessing and predicting small industrial enterprises’ credit ratings: A fuzzy decision-making approach
2022, International Journal of ForecastingReclassification of climatic zones for building thermal regulations based on thermoeconomic analysis: A case study of Turkey
2021, Energy and BuildingsCitation Excerpt :For this reason, the fuzzy clustering method is proposed since the data can belong to more than one cluster with different membership degrees [41]. The fuzzy c-means (FCM) algorithm, introduced by Dunn [42] and developed and extended by Bezdek [43] is a popular clustering approach and has been successfully applied in various areas including astronomy, geology, image segmentation, performance evaluation, power quality monitoring, risk analysis, tunnel construction, power systems, and region clustering [40,44,45]. In the FCM algorithm, each data provides membership to each cluster with a membership degree in the range of 0–1.
- 1
Tel.: +1 (508) 831 4831.