2021 : Improvement of Trip Distribution Model using Geographic and Spatial Weighting in Surabaya

Ir. Wahju Herijanto M.T.



Published in


External link






Until recently, the trip distribution model to be used with relatively high accuracy is the doubly constrained gravity model, but this model is still considered to have several weaknesses. The first is that it requires variables from the output of trip production and trip attraction modeling, the variables of which are the results of expensive field surveys or of those surveys carried out by the government, while those surveys are rarely available or published in many cities and areas. The second is that the coefficients obtained are related to the zone being modeled, so that it can predict the trip in the existing zone only; however, the trip prediction will not be applicable to the new extended zones (of the city). Meanwhile, other types of trip distribution model, the unconstrained gravity model, can be used for the development of new zones, but the accuracy is below the doubly constrained gravity model, while still using variables from published field surveys. The new data sources used as variables in trip distribution model in this study are digitation of settlements and places of activity from satellite imagery in Google Earth Pro, which is freely available, and also digitation of spatial which includes variations in urban spatial patterns to obtain six combinations consisting of three variations of city centers and two variations of the deterrence function. Geographical and spatial weights, in the form of coefficients and powers of equation, are obtained by optimization using the non-linear GRG method in the Solver facility of Microsoft Excel. The results show that all the six variations of the trip distribution model that combined geographical and spatial weights have less Normalized Mean Absolute Error (NMAE) than that of the conventional gravity models, which use population, employment, and student variables. One of the six models that assumes a concentric structure and a power deterrence function is also the most robust with almost equal NMAE when tested in cross validation. This particular model is a useful tool in measuring sustainable development goals (SDG) since it has better accuracy and transferability in urban transportation forecast with only one coefficient adjustment is needed.