PEMODELAN SPASIAL PERKEMBANGAN LAHAN TERBANGUN KOTA YOGYAKARTA DAN SEKITARNYA MENGGUNAKAN CELLULAR AUTOMATA DAN MULTI LAYER PERCEPTRON NEURAL NETWORK

Undirected urban growth can disturb comfort, so study about physical urban growth necessary to do. Urban growth is dynamic phenomena. Dynamic spatial modelling is suitable for this study, one of them of spatial modeling can be use is Cellular Automata (CA). This study use CA model that integrated with Multi Layer Perceptron Neural Network (MLP NN) to simulate physical urban growth of Yogyakarta City in 2009. Data used in this study are Landsat TM 1995 imagery and Landsat ETM 2002 imagery for basic data, then ALOS 2009 imagery for validation. The aim of this study are knowing Yogyakarta city growth from 1995 until 2002, simulating it in 2009, and knowing accuracy from simulation result. Yogyakarta city simulation in this study considering the road (local and main road) and activity center (CBD, university, and transportation infrastructure). Matrix area transition is required to makes the simulation. It shows possibility exchange of extents and transition to know distribution of possible changes. Matrix area transition obtained from Markov model and transition probability obtained from MLP NN model. Simulation is made with various scenarios which is based number of kernel and Ketetanggaan. Ketetanggaan used are Van Neuman and Moore Ketetanggaan with 3 x 3 and 5 x 5 kernel. The result of this study is Yogyakarta City Map in 2009 from simulation of various scenarios with an accuracy of 80.47% for Moore scenario (3x3 and 5x5), 70.30% for 3x3 Van Neuman scenario, and 70.69% for 5x5 Van Neuman scenario. Moore scenario in this study considered more precise because the result of accuracy is still tolerated. It happens because the number of Ketetanggaan that counted in Moore scenario more than Van Neuman scenario. Keyword : Cellular Automata, Multi Layer Perceptron Neural Network, Matriks area transisi, Probabilitas transisi, Neighbourhoo