Article Details

Abstract

The research object of this paper is the Fuzzy C-Means (FCM) clustering algorithm, which aims to improve the effectiveness of FCM in dealing with complex data. Traditional FCM assumes that the sum of the membership degree of a sample to each cluster is 1, which is called sample contribution degree. However, when dealing with noise points and outliers, the effectiveness of clustering may decrease. To this end, this paper proposes a new fuzzy clustering method. By relaxing the condition that the sum of membership degree is 1, a new membership degree division method is introduced, that is, the sum of membership degree is an innovative definition of the total number of samples n. This method dynamically adjusts the membership degree according to the distance between the sample and the clustering center, and divides the membership degree into three levels to improve the effectiveness of clustering. Through mathematical derivation and experimental verification, the improved algorithm improves the accuracy by 6.41% and the recall rate by 5.81% when dealing with complex data sets. It significantly improves the performance of the algorithm.

Keywords

FCM; membership degree; sample contribution degree; Target Detection

References

1. [1] Chen Qinan, Liang Hongjun Fuzzy and Rough Sets [J]. Computer Engineering, 2002, 28 (8): 138-140
2. [2] Wu Jing, Liu Yanheng, Lv Rong Distributed Learning Method Based on FCM [J]. Journal of Jilin University: Engineering Edition, 2010, 40 (1): 171-175
3. [3] HOPPER F，KLAWONN F. Improved Fuzzy Partitions for Fuzzy Regression Models [J].International Journal of Approximate Reasoning，2003，33（2-3）：85-102.
4. [4] Zhu Lin, Wang Shitong, Deng Zhaohong Generalized Study on FCM Clustering Algorithm with Improved Fuzzy Partition [J]. Computer Research and Development, 2009, 46 (5): 814-822
5. [5] Li Jie, Gao Xinbo, Jiao Licheng A new fuzzy clustering algorithm based on feature weighting [J]. Acta Sinica, 2006 (1): 89-92
6. [6] He Siyan, Li Peng, Liu Chengyu, etc Improvement and Influence Analysis of Membership Functions in Mutual Fuzzy Entropy [J]. Journal of Shandong University: Engineering Edition, 2014, 44 (1): 63-68
7. [7] Wei Yan, Li Xiaohong, Wu Xiao Fuzzy membership function weighted by posterior probability [J]. Journal of Chongqing University, 2012, 35 (8): 127-133
8. [8] Zou Peng, Yang Kaijun, Liang Chen Improving YOLOv5 for real-time detection of lightweight non-standard driving behavior [J]. Computer Engineering and Applications, 2023, 59 (13): 186-193
9. [9] Tan Jun Research on an Improved YOLOv3 Object Recognition Algorithm [D]. Wuhan: Huazhong University of Science and Technology, 2017
10. [10] TERVEN J，ESPARZA D M C，GONZÁLEZ J A R. A Comprehensive Review of YOLO Architectures in Computer Vision: From YOLOv1 to YOLOv8 and YOLO-NAS [J].Machine Learning and Knowledge Extraction，2023，5（4）：1680-1716.
11. [11] Li Bicheng, Zhang Chenxi, Ji Yuxiang, etc Accuracy Analysis of Artificial Intelligence Recognition and Artificial Recognition of Infrared Camera Animal Images: A Case Study of Macaque Monitoring on Dajinshan Island, Shanghai [J]. Journal of Ecology and Rural Environment, 2023, 39 (7): 918-923
12. [12] Lv Zongzhe, Xu Hui, Yang Xiao, etc YOLOv5 safety helmet detection algorithm for small targets [J]. Computer Applications, 2023, 43 (6): 1943-1949
13. [13] Zhou Lifang, Liu Jinlan, Li Weisheng, etc Twin Network Target Tracking Method Based on IoU Constraints [J]. Journal of Beihang University, 2022, 48 (8): 1390-1398

Funding

Funding from the "Blue and Green Plan" for universities in Jiangsu Province; 2024 Jiangsu Province Innovation Training General Funding Project (202413654018Y); 2023 Nanjing University of Science and Technology Zijin College Student Innovation Training Program Funding Project (ZJSRTP 2023118)

About the authors

Qian Siyuan (2004-), male, Han ethnicity, born in Nanjing, Jiangsu, is currently pursuing a bachelor's degree with research interests in software engineering, artificial intelligence, and big data.