基于改进支持向量机的瓦斯涌出量预测

为了克服瓦斯涌出量预测传统模型存在泛化能力弱和预测精度低的缺点，基于改进粒子群优化支持向量机建立一种非线性的煤矿瓦斯涌出量预测新模型。用改进的粒子群优化算法对支持向量机的惩罚因子与核参数进行寻优，选取最佳参数，以最佳参数对给定的训练样本进行学习训练，得到系统输入输出之间依赖关系的估计，再由这种关系对未知输出做出预测，进而建立起新型支持向量机预测模型。仿真实验结果显示，与普通粒子群优化的支持向量机相比，改进算法可使预测值的最大误差降低3．86％，平均误差降低4．27％，即新模型能够克服传统预测模型人为选取参数的盲目性以及神经网络的过学习问题，从而提高瓦斯涌出量预测的精度。

Forecasting of gas emission based on improved particle swarm optimizing support vector machine

The traditional forecasting model of gas emission ＇has the shortcomings of weak gener- alization ability and low accuracy of prediction. Based on improved particle swarm optimization, a kind of nonlinear support vector machine model of coal mine gas emission forecasting is built up in this paper. Using improved particle swarm optimization algorithm for support vector machine punishment factor and kernel parameter optimization, the best parameters are selected. By learn- ing with optimal parameters for the given training sample training, the dependencies between sys- tem input and output are estimated. They can be used to predict the unknown output as accurate- ly as possible. The forecasting model can then be built. Simulation results show that compared with ordinary support vector machines with Particle Swarm Optimization, the improved algorithm can reduce the maximum error of the predicted by 3.86% and mean error with reduced 4.27%. This suggests that the new model can overcome the blindness of traditional artificial selection pa- rameters and over-fitting of neural networks, and therefore can improve the precision of gas emis- sion forecasting effectively.