Arrival Place Presumption Based on Destination Entropy from Driving Route History

The relation between the destination entropy - the index showing the difficulty to estimate the arrival place - and the accuracy of the arrival place presumption from driving history was investigated. In previous methods for destination estimation from driving history, the presumption was based on the arrival place frequency in each driver's history. Therefore, it was greatly dependent on the content of the history runs as well as on each driver. In our method, the difficulty of presuming the arrival place was quantified by calculating the destination entropy from the driving history. In order to clarify the relation between the destination entropy and the performance of the proposed method, we calculated the entropy using real driving history of ten drivers. A strong correlation between the destination entropy and the accuracy of the presumed destination was shown. In the first five to ten minutes after the driving start we have obtained a significant entropy reduction and accuracy improvement. The destination entropy increased with the increase in the number of runs, but it converged after a certain predetermined value. The number of runs at which the convergence appears changes with each driver, and despite the increase of the accumulated driving history data, the presumption accuracy did not improve.