A Study of Fuzzy Clustering Ensemble Algorithm Focusing on Medical Data Analysis
Unitary clustering algorithm, not well adapted for fuzzy medical data sets, may result in low clustering accuracy and other problems. This paper investigates and compares the effects of various clustering methods to achieve improvements. First, unitary cl
- PDF / 809,370 Bytes
- 14 Pages / 439.37 x 666.142 pts Page_size
- 13 Downloads / 143 Views
Abstract Unitary clustering algorithm, not well adapted for fuzzy medical data sets, may result in low clustering accuracy and other problems. This paper investigates and compares the effects of various clustering methods to achieve improvements. First, unitary clustering algorithms such as k-means, FANNY, FCM, and etc. are achieved, then FCM algorithm was improved into CFCM algorithm, which increases the accuracy to a certain extent. Second, on this basis, in order to better adapt to the diversity of characteristics of fuzzy medical data, weighted co-association matrix is adopted to achieve integration, and consistency function is designed to present a fuzzy clustering ensemble algorithm. Finally, experiments shows that the Fuzzy Clustering Ensemble Algorithm can solve the problem of low accuracy in unitary clustering algorithm with higher stability, accuracy and robustness.
⋅
Keywords Medical data Fuzzy clustering ensemble algorithm tering Clustering ensemble
⋅
⋅
Fuzzy clus-
1 Introduction As known, discovering different types of diseases and classifying medical samples accurately are very important for successfully diagnosing and treating disease. For specific medical data, how to choose the appropriate clustering algorithm has always been the focus of the study. However, most of the studies are focusing on a unitary clustering algorithm for clustering medical data, which lacks robustness, stability and accuracy. However medical data has its unique characteristics, such as high-dimensional, vague and diverse. Therefore, it is important to carry out a comparative study on clustering algorithms for specific medical data. Z. Zhao (✉) ⋅ Y. Liu ⋅ J. Li ⋅ J. Wang ⋅ X. Wang School of Information Science and Engineering, Hebei North University, Zhangjiakou, China e-mail: [email protected] © Springer Nature Singapore Pte Ltd. 2018 N.Y. Yen and J.C. Hung (eds.), Frontier Computing, Lecture Notes in Electrical Engineering 422, DOI 10.1007/978-981-10-3187-8_37
383
384
Z. Zhao et al.
When clustering these data objects, any single type of attribute information is not enough to fully convey the data object. These different types of attribute information can complement each other and describe the entire data objects, thus it’s necessary to consider the integration problem among different types of feature during the clustering process. Traditional method weights different types of feature into an attribute vector to form a cluster when calculating the similarity of two data objects, resulting into local optimal solution under normal circumstances instead of the global one, which requires the development of efficient algorithm for data with high computational complexity. And many data samples do not have strict attributes, presenting betweenness in nature and generic terms. Under this circumstance, it is more suitable for soft partition to solve the fuzziness of data. The uncertainty of classification in fuzzy clustering illustrates the betweenness of the sample attributes, objectively reflects the diversity of clus
Data Loading...
