Visualizing crash data patterns

Abstract

This paper demonstrates an approach that makes it easy to find patterns in traffic crash data-bases, and to specify their statistical significance. The detected patterns might help to prevent traffic crashes from happening, since they may be used to tailor campaigns to the community at hand. Unfortunately, the approach described here comes at a cost: it identifies a considerable amount of patterns, not all of them are being useful. The second disadvantage is that is needs a certain size of the data-base: here it has been applied to a data-base of the city of Berlin that contains about 1.6 Million (M) crashes from the years 2001 to 2016, of which about 0.9M had been used in the analysis.

Keywords

• crash data patterns
• crash analysis
• contingency tables
• pearson residual
• cramers v
• mosaic plot