The soft soil subgrade settlement prediction is about a small sample,nonlinear and high dimensional data processed,support vector machine could solve this kind of problem. In order to overcome the theory and method in practical applications of penalty factor and kernel function parameter selection improper model prediction accuracy was not high,the use of particle swarm optimization algorithm to optimize the model parameters and PSO. Engineering examples showed that the PSO (Particle Swarm Optimization) support vector machine has higher accuracy,the prediction effect was better than that of PSO to optimize the support vector machine (SVM),also better than GRNN neural network and BP neural network prediction result,it was worth reference to engineering and technical personnel.%软土路基沉降预测是一个少样本、非线性、高维数据处理问题，支持向量机能够较好地解决这类问题。为了克服该理论方法在实际应用中存在惩罚因子C和核函数参数σ 选取不当导致模型预测精度不高的问题，采用粒子群优化算法 PSO对模型参数C和σ 进行优化。工程实例表明经 PSO优化的支持向量机具有较高的精确度，预测效果优于非 PSO优化的支持向量机，也优于GRNN网络和 BP神经网络的预测结果，值得工程技术人员借鉴。
展开▼
