Analysis of spatial autocorrelation for traffic accident data based on spatial decision tree

With rapid increase of scope, coverage and volume of geographic datasets, knowledge discovery from spatial data have drawn a lot of research interest for last few decades. Traditional analytical techniques cannot easily discover new, implicit patterns, and relationships that are hidden into geographic datasets. The principle of this work is to evaluate the performance of traditional and spatial data mining techniques for analysing spatial certainty, such as spatial autocorrelation. Analysis is done by classification technique, i.e. a Decision Tree (DT) based approach on a spatial diversity coefficient. ID3 (Iterative Dichotomiser 3) algorithm is used for building the conventional and spatial decision trees. A synthetically generated spatial accident dataset and real accident dataset are used for this purpose. The spatial DT (SDT) is found to be more significant in spatial decision making.