발간년도 : [2023]
| 논문정보 |
|
| 논문명(한글) |
[Vol.18, No.5] Addressing Issues in Fuzzy Clustering Using Neural Network Optimization Techniques |
|
| 논문투고자 |
Jong Chan Lee |
|
| 논문내용 |
Fuzzy clustering, a method representing the probabilities of belonging to k clusters using membership functions in the range of [0, 1], has been widely utilized. In this paper, we examine the issues that can arise during the utilization of this method from two perspectives. The first perspective introduces the challenge of classifying ambiguous data in fuzzy clustering and examines approaches to address it. For data that is hard to categorize, the method involves assigning it not forcefully to one cluster as a probability, but rather classifying it into a distinct cluster. The second aspect involves the application of fuzzy clustering in a field where it addresses the treatment of missing values in incomplete data. This problem originates from the notion that objects within arbitrary clusters obtained through fuzzy clustering share similar attributes, leading to the idea of estimating missing values from these clusters. The core of this paper lies in the development of an objective function that appropriately represents the presented problem and the application of an algorithm to optimize this objective function. The SA and EMFA utilized for optimization are applied by combining Boltzmann probability and annealing processes within the Monte Carlo method to increase the likelihood of approaching the global minimum. Through experimental processes, it is confirmed that these techniques produce favorable results. |
|
| 첨부논문 |
|
|
|
|
|
자료실 > 논문지 
