Lightning forecast based on the improved DBSCAN and polynomial fitting
Feng Wan-xin · Computer Engineering and Science · 2014
Using the improved DBSCAN algorithm considering the original characteristics of the lightning data,a lightning forecast method based on the improved DBSCAN and polynomial fitting is proposed to improve the prediction accuracy.Firstly,the lightning data during a span of time is clustered according to the density,and the average coordinates of all of the lightning data are designated as the central points of the cluster.Secondly,the central points in the last span of time are chosen as the initial selection points in the current span of time to perform DBSCAN clustering.The two-step procedure is iterated until all history data are processed.Finally,the polynomial fitting method is used to process the central location of each category.After dealing with the data provided by the Lightning Detection Network,the estimated and the observed performance data are presented and compared.The results are encouraging since the deviation between the actual data and the forecasted data is around 0.1 in most situations.