Land value is used for calculating Tax on Acquisition of Land Right and Building (BPHTB) and cost of land services like transfer of land rights. It is necessary for an accurate land value so that the calculation BPHTB and cost of land service is appropriate. The fact that the land value isn�t accurate, so we need a method/model to obtain land value accurately. A land value method that is suggested by the International Association of Assessing Officers (IAAO) is Artificial Neural Network (ANN). The high transfer of land rights in the Trihanggo Village certainly needed land value accurately. Therefore, this research aims to determine: 1) the stages for modeling land value by using ANN model and 2) land value modeling results by using ANN model in the Trihanggo Village,Gamping sub district, Sleman District of Yogyakarta.
This research used a survey method with quantitative approach. The technique of collecting data through interviews and document study. Sampling was done by purposive sampling. Sample count 85 samples of land parcels that have been done transactions of sale and supply. Variables suspected to affect land values include the land area (LT); front width (LD); land use (PGT); land shape (BB); land position (LH); land right (SK); class of road (KJ); distance to arterial roads (JA); distance to collector road (JK); distance to healthcare (JR); distance to market (JP) and distance to river (JS). For determining the variables that significantly affect to land value is done by variable selection based on public
perception and statistical analysis. To get the best ANN model is tested of 8 training algorithm. The best ANN models is evaluated by looking at the level of accuracy (Coefficient of Variation/COV), the level of uniformity (Coefficient of
Dispersion/COD) and the level of fairness (Price Related Differential/PRD).
The results showed that: 1) ANN can be applied for modeling land value with steps by variable selection, network architecture design and perform the model of ANN; 2) The best ANN model based on the public perception is Backpropagaion Resilent training algorithm and the best ANN model based on statistical analysis is One Step secant training algorithm; 3) Evaluation of land value model by using ANN model based on the public perception has COV value of 12.17%, COD value of 14.41% and PRD value of 1.00 while the ANN models based on statistical analysis has COV value of 15.74%, COD value of 17.68% and the PRD value of 1.02.

Keywords: Land Value Model, Artificial Neural Network