Research on MDS and Semi-supervised Clustering Algorithm
Dong Chen, Changxin Song · 2021 IEEE International Conference on Computer Science, Electronic Information Engineering and Intelligent Control Technology (CEI) · 2021
With the rapid development of computer technology, the dimensions of data have exploded, and many data analyses have become very difficult. To solve the above problems, the use of dimension reduction method of the multidimensional scaling transformation (MDS) for data dimension reduction, get rid of some redundant information, save the storage space. In view of the traditional k-means clustering method, by introducing a pair constraint supervision information to guide the clustering process, formed a semi-supervised k means clustering approach. In addition, on the basis of the Cop-Kmeans algorithm based on Breadth-first search (BFS), I design an improvement based on data segmentation technology in terms of the selection of the initial clustering center. In this paper, MDS and the improved$\text{BFS}+\text{Cop}$-Kmeans algorithm are fused to form the$\text{MDS}+\text{BCK}$algorithm, and on the UCI data set an experimental verification was performed. The first results of the test indicate that the performance of the present algorithm shows further improvements compared to the traditional k-mean algorithm, the Cop-Kmean means, and the unimproved$\text{BFS}\ +\ \text{Cop}$- Kmean algorithm.