Shallow landslide prediction using a novel hybrid functional machine learning algorithm

摘要: Coastal wetland mapping plays an essential role in monitoring climate change, the hydrological cycle, and water resources. In this study, a novel classification framework based on gravitational optimized multilayer perceptron classifier extended multi-attribute profiles (EMAPs) is presented for coastal using Sentinel-2 multispectral instrument (MSI) imagery. proposed method, morphological attribute (APs) are firstly extracted four filters characteristics of wetlands each band from These APs form set EMAPs which comprehensively represent irregular objects multiscale multilevel. The original spectral features then classified with new (MLP) whose parameters by stability-constrained adaptive alpha search algorithm. performance method was investigated MSI images two wetlands, i.e., Jiaozhou Bay Yellow River Delta Shandong province eastern China. Comparisons other classifiers through visual inspection quantitative evaluation verified superiority method. Furthermore, effectiveness different were also validated. By combining developed MLP classifier, complicated types high within-class variability low between-class disparity effectively discriminated. superior makes it available preferable data similar optical