Improved SOM based data mining of seasonal flu in mainland China
Tao Xu, Jieping Zhou, Jianhua Gong, Wenyi Sun, Li‐Qun Fang, Yanli Li · 2012
SOM (Self-Organizing Maps) is an efficient and unsupervised data mining method based on neural networks clustering algorithm. It has become a new hotspot in data mining and visualization. Data mining methods are often manipulated by epidemiologists to reveal the regulation of the diseases. As the temporal characteristics exist in the data of the infectious diseases, and the limitations of the original SOM method, this paper introduce an improved position adjustable SOM and then study seasonal flu in mainland China in 2006 with it. The result shows the improved SOM is an efficient method analyses spatiotemporal infectious disease data. Using this improved SOM method, some abnormal data in dataset has been easily found, such as data-entry error. Clustering result revealed several types of flu virus transmitted in different Chinese cities in 2006, which include: winter flu peaks in the Yangtze River delta areas, winter and summer flu peaks in Donggang, and the whole year high rate of flu in Tianjing, Honghe and Guilin.