关键词: 可视化, 聚类分析, 带约束的优化, 遗传算法

Abstract: This paper validated that the Cluster Constrained Mapping (CCM) can keep the "topological" information of the points in the reduced dimension map by comparing the cluster results obtained using the K-means algorithm. The enhanced GA proposed in Part I was applied to three constrained optimization cases. The results show that the combination of visualization, cluster analysis and genetic algorithms can help users to participate in selecting appropriate parameters of clusters, and the combination of a computer and the user is more powerful than either alone, which is an effective process optimal design tool with high solution quality and consistency. In the new cluster analysis method, the data are visualized by CCM that provides immediate direct information about the feasible domain, and the user is directly involved in determining the parameters for the cluster analysis and increasing the effectiveness of feasible regions discovery by visual interaction; the obtained knowledge is visualized by Parallel Coordinate Systems (PCS), thus the user has a deeper understanding of the feasible regions. It is clear that in most cases the proposed IGA based on the combination of visualization and cluster analysis has performed not only with the high efficiency (in terms of getting closer to the best-known solution) and with more robustness (in terms of the number of GA runs finding solutions close to the best known solution), but also with providing more information about the feasible regions for the user to understand the model and accept the optimal results.

Key words: visualization, cluster analysis, constrained optimization, genetic algorith