Page 800 - NGTU_paper_withoutVideo

Modern Geomatics Technologies and Applications

Parks p31=0.008 p32=0.006 p33=0.031 p34=0.022 p3t=0.067
Agricultural
lands p41=0.072 p42=0.004 p43=0.009 p44=0.107 p4t=0.192
Total pt1=0.349 pt2=0.392 pt3=0.067 pt4=0.192 1

Table 5. Number of changeable pixels per user for the years 2002 to 2008 in SimWeight method.

Use Changeable area Number of changeable pixels

Agricultural lands 3049.46 34185
Buildings 55497.58 67528
Parks 12467.28 14795
Wastelands 40153.84 45729

(a) (b)

Fig. 6. Prediction map with SimWeight method (a) soft prediction 2014 map, (b) hard prediction 2014 map.

6. Conclusion

In this paper, Similarity Weighted Instance-based method was used to model the land use changes. Kappa coefficient and
overall accuracy indicate the high capacity and power of this method for modelling densely populated urban areas. The results
of urban development simulation showed a relatively good accuracy of the SimWeight model (see Table 6) for the urban area
with 4 land uses. Therefore, it can be recommended to the relevant decision makers and for predicting urban development.

Table 6. Results of model validation, SimWeight using matching simulated map with earth reality map.

Overall accuracy Kappa coefficient

81.46 87.23 SimWeight method

7. Reference

[1] C. Mozumder, N. K. Tripathi, and C. Losiri, “Comparing three transition potential models: A case study of built-up
9