A Novel Triangulate Mapping Based on Self-Organized Anchor Points for Data Visualization
Ming Leong Yii, Chee Siong Teh · Advanced Science Letters · 2017
Without a form of visual feedback, multivariate data would be reduced to a lump of numbers that very few people would be able to appreciate and be benefited from. This research paper proposes a novel triangulate mapping technique based on selforganizing anchor points for multivariate data visualization. Self-Organizing Map (SOM) and a modified Adaptive Coordinates (AC) are hybridized to produce the anchor points in the 2D space. The trained anchor points are used to triangulate data onto a topologically preserved 2D space. The empirical studies that produce topologically preserved data visualizations for high dimension and arbitrarily shaped clusters in simulated, benchmarking, and real-life dataset show its usefulness in providing intuitive visual feedback to the user.