论文部分内容阅读
为了对公路运行车速进行准确预测,采集了国道二级公路上232处典型路段的平曲线半径和纵坡度等线形数据和小轿车车速,分别应用线性回归、多项式回归和智能学习型方法中的BP神经网络和模糊神经网络建立了小轿车第85百分位车速模型,并对4种模型预测精度进行了对比分析。结果表明:在回归模型中,多项式回归的预测精度优于线性回归;在学习型算法中,模糊神经网络的预测精度优于BP神经网络,并且模糊神经网络的预测精度优于统计回归方法;从模型使用角度来看,相比较于线性回归模型要求样本随机误差满足零均值和正态分布等假设条件,神经网络算法的限制条件较低,具有更广阔的应用前景。
In order to accurately predict the running speed of the highway, the linear data such as the radius of the curve and the longitudinal gradient of the 232 typical sections of the national highway are collected and the speed of the car is collected. The linear regression, polynomial regression and BP Neural network and fuzzy neural network established the 85th percentile vehicle speed model, and compared the prediction precision of 4 kinds of models. The results show that the prediction accuracy of polynomial regression is better than that of linear regression in regression model. In the learning algorithm, the prediction accuracy of fuzzy neural network is better than that of BP neural network, and the prediction accuracy of fuzzy neural network is better than that of statistical regression. Compared with the linear regression model, the random error of the sample meets the assumptions of zero mean and normal distribution. The restriction of the neural network algorithm is lower and has a broader application prospect.