One form of Automated Valuation Models (AVM) in mass land valuation is Artificial Neural Network (ANN) model. However, ANN model provided in the form of algorithm is too complex to be implemented in Land Office. Therefore, this study aims to develop application of mass land valuation based on neural network using Graphical User Interface (GUI) for easy use and accurate land mass assessment in parcel based to produce ZNT map. This research using Research and Development method. Data were collected using interviews, direct observations and literature study methods. Sampling was done by multistage random sampling method, by choosing 102 sample parcels which have been traded or on dealing process. The variables considered to affect land value are (1) the extent of the land (LTH); (2) the width of the field (LD); (3) the shapes (BT); (4) land use (PGT); (5) the location (LH); (6) the status of ownership (SH); (7) road function (FJ); (8) distance to arterial road (JA); (9) distance to collector road (JK); and (10) distance to health facilities (JRS). ANN algorithm used as the basis for the development of applications of was backpropagation gradient descent, while the software used for GUI development was Matlab.
The result of this research is an application of mass land valuation, referred as ANNAVAL (Artificial Neural Network for Mass Valuation). ANNAVAL testing was done by mass land assessment using 6 different parameter setting conditions. Validation test showed that land value in Condition 5 has the smallest transforming factor, which is 13.2%. Validation test was done by comparing land value obtained from ANNAVAL calculation with land value from validation result in the same parcels. This study is expected to provide useful software for Land Offices to perform parcel based � mass land valuation, that can be easily and quickly operated; besides provide accurate land value datas for ZNT map.
Keyword: models, applications, land valuation, land value, artificial neural network, matlab