PSO—SVR在果酒生物活性物质预测中的应用

针对BP神经网络和遗传算法对果酒生物活性物质预测存在速度慢和精度低的缺点，建立了基于支持向量回归机（SVR）的果酒生物活性物质预测模型。鉴于支持向量机模型的精度和泛化能力很大程度取决于不敏感损失系数ε、惩罚系数C和RBF核函数的宽度系数7三个参数，模型采用粒子群算法对三个参数同时进行优化，实现了果酒生物活性物质的非线性预测。仿真结果表明：基于PSO—SVR算法的果酒生物活性物质预测模型性能优于所比较的BP神经网络模型和支持向量回归机模型，能有效提高果酒生物活性物质的预测精度和稳定性。

Prediction Model of Bioactive Substances From Wine Based on PSO-SVR

To solve the predict presence of slow speed and low precision of the bioactive substances from wine by using BP neural network and Genetic Algorithm, a prediction model based on Support Vector Regression （SVR）has been designed for metal corrosion rate. It is well known that the model complexity and generalization performance of this Support Vector Regression model depend on setting of the three parameters （8, C, ~/）. Using the algorithm called Particle Swarm Optimiza- tion （PSO） to optimize the three parameters at the same time, nonlinear predictive of the bioactive substances from wine was achieved. Simulation results show that the proposed model is superior to the other two models for improving the forecast accuracy and stability.