【摘要】 针对灰色(1,1)模型(Grey model(1,1), GM(1,1))对非指数型数据序列预测精度低的问题,提出了一种灰色支持向量回归(Grey support vector regression, GSVR)预测模型。该模型一方面通过参数累积估计、预测公式改进和数据等维递补,对灰色模型进行建模优化,另一方面通过差分变异和混沌局部搜索改进的粒子群算法,对支持向量回归机进行参数优化,再将二者相结合进行预测。对柱塞套内圆珩磨尺寸的预测结果表明,该模型的预测均方误差为0.3913,平均绝对百分比误差为4.90%,其预测精度较GM(1,1)模型显著提高。更多还原

【Abstract】 To improve the forecasting accuracy of grey model(1,1)(GM(1,1)) for non-exponential data, a grey support vector regression(GSVR) model was proposed in this paper. On the one hand, the GM(1,1) was optimized by parameter accumulation estimation, forecasting formula deduction and data equal dimension complement; On the other hand, the support vector regression(SVR) was optimized by the particle swarm optimization(PSO) which was improved by differential mutation and chaotic local search; then the GSVR model was obtained by combining the optimized GM(1,1) and SVR. The forecasting results of the plunger bushing internal honing size suggest that the mean square error(MSE) of the proposed model is 0.3913, the mean absolute percentage error(MAPE) is 4.90%, and the forecasting accuracy is significantly enhanced compared to GM(1,1).更多还原