Crack Detection using Spectral Clustering: Self-Tuning Considering Crack Feature and Connections

Shiotsuka, Daiki; Matsushima, Kousuke; Takahashi, Osamu

doi:10.1109/morse48060.2019.8998708

Cited by 1 publication

(1 citation statement)

References 18 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…e number of eigenvalues is determined by the number of clustering groups after obtaining the eigenvectors. Shiotsuka et al [22] proposed the Mcut algorithm. is algorithm deeply studies the vector and uses this vector as a criterion as a feature vector.…”

Section: Related Knowledgementioning

confidence: 99%

[Retracted] Audit Analysis of Abnormal Behavior of Social Security Fund Based on Adaptive Spectral Clustering Algorithm

Chen

Wen-hao

2021

Complexity

View full text Add to dashboard Cite

Abnormal behavior detection of social security funds is a method to analyze large-scale data and find abnormal behavior. Although many methods based on spectral clustering have achieved many good results in the practical application of clustering, the research on the spectral clustering algorithm is still in the early stage of development. Many existing algorithms are very sensitive to clustering parameters, especially scale parameters, and need to manually input the number of clustering. Therefore, a density-sensitive similarity measure is introduced in this paper, which is obtained by introducing new parameters to transform the Gaussian function. Under this metric, the distance between data points belonging to different classes will be effectively amplified, while the distance between data points belonging to the same class will be reduced, and finally, the distribution of data will be effectively clustered. At the same time, the idea of Eigen gap is introduced into the spectral clustering algorithm, and the verified gap sequence is constructed on the basis of Laplace matrix, so as to solve the problem of the number of initial clustering. The strong global search ability of artificial bee colony algorithm is used to make up for the shortcoming of spectral clustering algorithm that is easy to fall into local optimal. The experimental results show that the adaptive spectral clustering algorithm can better identify the initial clustering center, perform more effective clustering, and detect abnormal behavior more accurately.

show abstract

Section: Related Knowledgementioning

confidence: 99%