Knowledge transfer and adaptation for urban simulation cellular automata model base on multi-source TrAdaBoost algorithm

Traditional cellular automata (CA) cannot adequately simulate urban dynamics and land-use changes when there are insufficient training samples. To address this problem, we propose a multi-source knowledge transfer CA model. This model utilizes several existing label data sets to help train a new model. This proposed model, MSTra CA, is employed to urban simulation in Shenzhen City in Guangdong Province of China. Experiments have demonstrated that the proposed method can alleviate the sparse data problem using knowledge transfer thus reducing the negative transfer learning risk.