Data mining for the advection database of wake vortices

Advection database of wake vortices were constructed based on the Doppler lidar measurements at Sendai Airport to understand the relations among measurement, weather factors and behavior of wake vortex. The advection database of wake vortex consists of three measurement factors and fifteen weather factors as variables to determine the wake vortex behavior. However, the extraction of correlations from the high-dimensional data is difficult in general. Therefore, in this paper, data mining was employed as an efficient approach to extract relations among the variables from the advection database of wake vortex. As a first step of the data mining, redundant measurement, weather factors were removed through Spearman rank method, then Self-Organizing Map (SOM) was created to find correlations among the measurement, weather factors and wake vortex behavior. Next, the measurement, weather factors which have larger influences to the behavior of the wake vortex were identified by Analysis of Variance (ANOVA). Finally, rough set theory revealed the specific rules of measurement, weather factors to determine wake vortex behavior.