A modular artificial neural network valuation system
Panayiotis Panayiotou, Constantinos S. Pattichis, David H. Jenkins, Frances Plimmer · 2005
The objective of this study was to design, develop and test a modular artificial neural network valuation system (MANN) incorporating multiple regression analysis (MRA) for the assessment of houses and apartments in Strovolos municipality in Cyprus for taxation purposes. The system combined the assessment results of three neural network assessors: (i) back-propagation (BP), (ii) probabilistic network (PNN) (iii) self-organising feature map (SOFM); and (iv) MRA. Features include age, size, plot size, date of sale etc. The mean absolute percentage error (MAPE) and the coefficient of dispersion (COD) of the MANN system for houses were 10.67% and 10.57% respectively. The MAPE and the COD for apartments were 8.68% and 8.41% respectively. These findings compare favourably with other studies and satisfy International Association of Assessing Officers (IAAO) criteria for mass appraisal techniques.