Qihang GuoVerified
Verified Account
Verified
Artificial Intelligence | Machine Learning | Medicine
Member for 12 days
Research Achievements: Authored numerous publications in top-tier journals within Computer and Systems Sciences, including Pattern Recognition (PR), Expert Systems with Applications (ESWA), Information Sciences (INS), and Engineering Applications of Artificial Intelligence (EAAI), among others.
Achievements
Cited Author
Peer Reviewer
Key Stats
Upvotes received:
0
Publications:
14
(21% Open Access)
Cited by:
153
h-index:
6
/
i10-index:
3
Reputation
Organic Chemistry
64%
Artificial Intelligence
36%
Computational Theory And Mathematics
17%
Show more
How is this calculated?
Publications
2

An Improved Three-Way K-Means Algorithm by Optimizing Cluster Centers

Qihang Guo et al.Sep 2, 2022
Most of data set can be represented in an asymmetric matrix. How to mine the uncertain information from the matrix is the primary task of data processing. As a typical unsupervised learning method, three-way k-means clustering algorithm uses core region and fringe region to represent clusters, which can effectively deal with the problem of inaccurate decision-making caused by inaccurate information or insufficient data. However, same with k-means algorithm, three-way k-means also has the problems that the clustering results are dependent on the random selection of clustering centers and easy to fall into the problem of local optimization. In order to solve this problem, this paper presents an improved three-way k-means algorithm by integrating ant colony algorithm and three-way k-means. Through using the random probability selection strategy and the positive and negative feedback mechanism of pheromone in ant colony algorithm, the sensitivity of the three k-means clustering algorithms to the initial clustering center is optimized through continuous updating iterations, so as to avoid the clustering results easily falling into local optimization. Dynamically adjust the weights of the core domain and the boundary domain to avoid the influence of artificially set parameters on the clustering results. The experiments on UCI data sets show that the proposed algorithm can improve the performances of three-way k-means clustering results and is effective in revealing cluster structures.
Load More