基于杂交粒子群算法的多元线性回归参数估计及预测区间研究

为了提高多元线性回归分析模型预测混凝土28 d抗压强度的准确性和可靠性,采用杂交粒子群优化算法估算模型系数,依据正态分布和t分布求出预测点置信度为95%的预测区间。实验结果表明:此模型的预测精度优于传统基于最小二乘估算的回归分析模型,且预测结果可以是相关量的取值范围,扩大了相关量的适用范围,提高了预测的可靠性。

Multiple linear regression parameters estimation and prediction intervals research based on hybrid particle swarm optimization

In order to improve the accuracy and reliability of multivariate linear regression analysis model in predicting the 28d com- pressive strength of concrete, using hybrid particle swarm optimization algorithm to estimate the model coefficients, according to normal distribution and t distribution, the prediction interval with predicted point of 95% confidence was found out. The experiment results show that the prediction accuracy of the model is superior to that of the traditional regression analysis model based on the least square estimation, and the predicted results can be the value range of related parameters, which expands their application scope and improves the prediction reliability.