Sediment Carrying Capacity Prediction Based on Chaos Optimization Support Vector Machines

Correct calculation of sediment carrying capacity in natural rivers is of great significance to the simulation of sediment movement and river-bed deformation by mathematical model. Peak recognition support vector machines, an improved support vector machines, was proposed considering the complication and nonlinearity between sediment carrying capacity and its impact factors; peak recognition least square support vector machines sediment carrying capacity prediction model, which was based on chaos optimization, was built combining with accelerating chaos optimization against questions of support vector machines regression such as parameter optimization, training and test speed. The test data of 30 sets of water tanks with high, medium and low sediment concentrations were trained, and training values agreed well with measured values; four sets of test data were predicted by trained support vector machines model, and training values were pretty much the same with measured values. Theoretical analysis and experimental results show that sediment carrying capacity studying method based on peak recognition support vector machines is more accurate in predication and more reliable than common support vector machines and BP neural network.