Volume XLII-4/W18
Int. Arch. Photogramm. Remote Sens. Spatial Inf. Sci., XLII-4/W18, 1071–1078, 2019
https://doi.org/10.5194/isprs-archives-XLII-4-W18-1071-2019
© Author(s) 2019. This work is distributed under
the Creative Commons Attribution 4.0 License.
Int. Arch. Photogramm. Remote Sens. Spatial Inf. Sci., XLII-4/W18, 1071–1078, 2019
https://doi.org/10.5194/isprs-archives-XLII-4-W18-1071-2019
© Author(s) 2019. This work is distributed under
the Creative Commons Attribution 4.0 License.

  19 Oct 2019

19 Oct 2019

EXPLOITATION OF MCDA TO LEARN THE RADIAL BASE NEURAL NETWORK (RBFNN) AIM PHYSICAL AND SOCIAL VULNERABILITY ANALYSIS VERSUS THE EARTHQUAKE (CASE STUDY: SANANDAJ CITY, IRAN)

P. Yariyan1, M. R. Karami2, and R. Ali Abbaspour3 P. Yariyan et al.
  • 1Dept. of Geography Information System, Mamaghan branch, Islamic Azad University, Mamaghan, Iran
  • 2Dept. of Social Science, Faculty of Humanities and Social Sciences, Payame Noor University, Tehran, Iran
  • 3Dept. of Surveying and Geospatial Engineering, University of Tehran, Iran

Keywords: Earthquake, MCDA, ANN, RBFNN, Sanandaj, IRAN

Abstract. Despite years of research on natural hazards and efforts to reduce physical and psychological damage, earthquake as a natural disaster is catastrophic. Though, human is the main axis in dealing with crisis and vulnerability, and since the space of cities encompasses largest population spectrum, managing this space is considered as an essential issue. Accordingly, the vulnerability of the City of Sanandaj was defined by environmental, physical and social criteria. In this regard, with the aim of modeling, and assessing the risk and vulnerability, the MCDA-ANN hybrid model was introduced as a new method for teaching of learning models. In order to determine the final value of each of the criteria, AHP analysis was performed as one of the MCDA methods to solve complex and non-structural problems by creating a functional hierarchy, and after that, a training data base for learning ANN was created randomly based on the AHP classification map. Then, for modeling, the radial base functional neural network (RBFNN) was used as one of the techniques of artificial neural networks. After the modeling, 30% of the points were selected as validation data to determine the accuracy of the model. After the implementation of RBFNN model, the area of AUC curve resulted is 0.922, which indicates the high accuracy of the model in assessing the risk of an earthquake. The results show high vulnerability in urban areas1 and 2 and in downtown Sanandaj that in these zones the physical and social factors dramatically affect the vulnerability of these areas.