LEE- Endüstri Mühendisliği-Doktora
Bu koleksiyon için kalıcı URI
Gözat
Konu "Çok algılayıcı veri füzyonu" ile LEE- Endüstri Mühendisliği-Doktora'a göz atma
Sayfa başına sonuç
Sıralama Seçenekleri
-
ÖgeDeriving weights of decision makers in group decision making and applications in medical decision making and sensor fusion( 2020) Köksalmış, Emrah ; Kabak, Özgür ; 636295 ; Endüstri Mühendisliği Ana Bilim DalıThe motivation behind the rational decision-making method is to determine the most proper alternative(s) from a set of alternatives regarding the predefined criteria. A structured and reasonable decision making process is essential to settling on rational and appropriate decisions. Especially, the use of rational approaches instead of subjective techniques stimulates organizations to take the correct decisions and cope with any difficulties, efficiently. Consequently, decision making methods have been applied efficiently in a variety of complex areas, such as the military, economics, government organization, and are increasingly attracting the attention of academics for years. Quality of the solution of the decision making depends fundamentally on the nature of the problem, but mostly on the characteristics of decision makers. As the complication of the socio-economic environment increases, it gets more problematic for single decision maker to handle all the relevant features of the problem. Most decision making problems in real world occur in a group environment and this adds too much complexity to the analysis. Therefore, academics are searching for appropriate group decision making (GDM) approaches in recent years to overcome this problem. GDM is a method in which a group of experts (i.e. decision makers, group members, voters, stakeholders) are gathered to find out the solution of the decision making problem. In this process, motivation and understanding of a common problem differ from one decision maker to other depending on the knowledge, background, and expertise of these decision makers. At this point, different weights can be assigned to these people reflecting their importance or perceived reliability for the given problem. In GDM problems, experts describe their preferences by taking each criterion into account, and final decision is reached by merging all decision matrices into an aggregated solution applying a proper operator. At this point, it is important to develop a better technique for aggregating different decision makers' preferences to obtain an acceptable decision making result. In the literature, GDM methods commonly assume that the decision makers have same level of importance weights and disregards the relative weights. This situation may cause inappropriate and inaccurate outcomes that cannot be compensated in the final result. Consequently, reliability and the significance of decision makers on the final decision should be taken into consideration. At that point, how to derive the appropriate weights of decision makers stands as a new challenge. Same challenge is also valid for the multi-source fusion problems that effort to find an appropriate technique to combine the data from multiple sources; for example, sensors, where each sensor may have different features. The key difference here is that the sensors, which may differ in specifications, are replaced with the decision makers whose expertise, background, or knowledge may also vary. Therefore, methods, which are developed to overcome this challenge, have several applications in wireless sensor networks, image fusion, etc. In literature, researches on deriving the weight of decision makers are relatively limited. Moreover, a comprehensive literature review on determining the weight of decision makers is missing among a limited number of studies. Consequently, in the second chapter of the thesis, the literature on deriving the objective weights of decision makers is studied and a new scheme for classification is proposed. According to the stated classification scheme; objective methods are divided into five groups: Similarity-based approaches, index-based approaches, cluster-based approaches, integrated approaches, and other approaches. Literature review and analysis of the studies in literature were conducted with respect to these categories. In the third chapter of the thesis; in order to demonstrate the application of integrated approaches, a new method, that derives decision makers' combined weights using the geometric weights consensus index (objective method) and the subjective weights provided by a supervisor, is developed. The application of the method is verified on a case study in a medical decision making problem, specifically, selection of a suitable anesthesia method to apply in the surgery which involves three alternatives such as the general anesthesia, local anesthesia and sedation. In the fourth chapter, a large scale GDM approach is proposed for the sensor fusion. Since the proposed method is a cluster-based method, it provides acceptable results in sensor networks consisting of multiple sensors. The method can operate under uncertainty as a result of converting raw data from sensors into basic probability assignments. In addition, by assigning three objective weights, the reliability of the sensor clusters was also taken into account. In addition to these objective weights, the proposed method allows subjective weights to be allocated to integrate the experience and knowledge of supervisors into the problem area. The applicability and validity of the proposed method have been checked with two real classification data sets. Experiments show that when the proposed method is applied to two data sets, the classification rate increases significantly. In the last part of the study, the effect of the expansion parameter, objective weights, reliability threshold, number of clusters and clustering method on the classification rate and probability of detection are examined. In the last chapter of the thesis, the results obtained from these studies, problem areas, limitations and potential research directions are discussed.