The role of visualization in effective data cleaning

Using visualization techniques to assist conventional data mining tasks has attracted considerable interest in recent years. This paper addresses a challenging issue in the use of visualization for data mining: choosing appropriate parameters for spatial data cleaning methods. On one hand, algorithm performance is improved through visualization. On the other hand, characteristics and properties of methods and features of data are visualized as feedbacks to the user. A 3-D visualization model, called Waterfall, is proposed to assist spatial data cleaning in four important aspects: dimension-independent data visualization, visualization of data quality, algorithm parameter selection, and measurement of noise removing methods on parameter sensitiveness. Keywords Information visualization, data mining, clustering, noise removal.

• Type: Conference
• Year: 2005
• Where: SAC
• Authors: Yu Qian, Kang Zhang
• Keywords: Applied Computing, Data Mining, Data Mining Tasks, SAC 2005, Spatial Data Cleaning