利用PSO-GA-LSSVM模型预测基坑周边建筑物沉降  

Prediction of Building Settlement Around Foundation Pit Using PSO-GA-LSSVM Model

作　　者：谢洋洋 付超[1] 吴大鹏[1] 卞晓晨 王春 XIE Yangyang;FU Chao;WU Dapeng;BIAN Xiaochen;WANG Chun(Provincial Geomatics Center of Jiangsu,Nanjing 210013,China)

机构地区：[1]江苏省基础地理信息中心,江苏南京210013

出　　处：《测绘地理信息》2021年第3期50-54,共5页Journal of Geomatics

基　　金：辽宁省教育厅重点实验室基础研究项目(LJZS001)。

摘　　要：针对最小二乘支持向量机(least square support vector machine, LSSVM)模型参数选择存在随机性与单一优化算法寻找参数存在局限的问题,将遗传算法(genetic algorithm, GA)、粒子群优化(particle swarm optimization, PSO)算法引入LSSVM模型,建立了基于粒子群-遗传算法(PSO-GA)优化的LSSVM沉降预测模型。将GA嵌入PSO算法,降低了模型参数寻优陷入局部最优的可能,提高模型拟合精度。结合具体工程实例,将提出的模型与LSSVM模型、PSO算法优化的LSSVM(PSO-LSSVM)模型、GA优化的LSSVM(GA-LSSVM)模型进行对比,结果表明改进模型的精度更好,稳定性更强。Aiming at the problems of parameter selection randomness in least squares support vector machine(LSSVM) model and the limitation of parameter searching in single optimization algorithm, the genetic algorithm(GA) and particle swarm optimization(PSO) algorithm are introduced into the LSSVM model, and then the optimized LSSVM settlement prediction model based on PSO-GA is established. The GA is embedded into the PSO algorithm to reduce the possibility of the model parameter optimization falling into local optimum and to improve the efficiency of model parameter optimization. Combined with the specific engineering examples, the proposed model is compared with the LSSVM model, PSO-LSSVM model and GA-LSSVM model. The results show that the improved model has better accuracy and stronger stability than the other three models.

关 键 词：最小二乘支持向量机 粒子群优化算法 遗传算法 粒子群-遗传算法 沉降预测 

分 类 号：TU196.2[建筑科学—建筑理论]