The built-up areas of 194 cities and towns (> 10 000 in population) in 2020 were taken as positive samples, and the buffer ranges of 400 non-urban random points generated in areas where slope > 10° and distance to river > 10 km as negative samples. Positive and negative samples were divided into a training set and a validation set in a ratio of 7:3 by random selection, and 300 classification trees were set within the RF model to classify the whole study area. The results were verified through the validation set and an overall accuracy (OA) of 94.88% and a Kappa Coefficient (KC) of 89.03% were obtained. This means that the prediction of the optimal urbanization candidate areas seems reliable. These areas are distributed in basins and valleys around the Poyang Lake, along the rivers Yangtze, Ganjiang, Xinjiang, Yuanhe, Jinjiang and Fuhe, and those along the deep and large fault zones are the most suitable ones for urban development. |
ACCESS THE FULL ARTICLE
No SPIE Account? Create one
Modeling
Algorithm development
Machine learning
Random forests
Artificial neural networks
Data modeling
Decision trees