Comparative Assessment on Regional and Trading Value of Residential Properties Estimating Using Artificial Neural Networks, Case Study: District 2 of Tabriz

Municipalities in Iran that serve as local governments in urban affairs must use local financial resources to cover their costs. One of these local financial resources is the cost of a building. Since the basis for calculating the charges have been the value of a real-estate transaction, the correct estimate has been very important. The aim of this paper is to use artificial neural network model for estimating residential property prices in district 2 of Tabriz; to achieve the regional value that is the basis for calculating the trading value. The approach of this research is applied and development researches. Statistical population is the residential property in district two of Tabriz which is 24638. Cochran formula was used to estimate the sample size and estimate 378 as samples. For desirable estimation 400 units were randomly selected. To remove the effect of time, only data from June to August 2018 were used. The data were collected through survey and inquiry from real-estate agents. In this research, were used MATLAB 2013 and ArcMap 10.4. The results show the high accuracy of the artificial neural network in estimating property prices. The trading value approved by all the blocks in district two of Tabriz is lower than the estimated value of the artificial neural network. The highest difference in estimation is in 13, 7 and 24 blocks (respectively -3050380, -2752550 and -2430850 Rial) and the lowest difference is in 9 block (-399850 Rial). Thus, the total trading value is “11,056,920” Rials less than the estimated value.