针对新冠肺炎疫情发展模型的预测问题,提出了一种基于改进遗传算法误差反向传播(genetic algorithms back propagation,GABP)神经网络算法。该算法通过对遗传算法中的交叉、变异、及适应度等函数进行了重新定义,在获得GABP神经网络最优...针对新冠肺炎疫情发展模型的预测问题,提出了一种基于改进遗传算法误差反向传播(genetic algorithms back propagation,GABP)神经网络算法。该算法通过对遗传算法中的交叉、变异、及适应度等函数进行了重新定义,在获得GABP神经网络最优参数的同时,提高了最优参数的泛化能力,进而缩小了GABP网络寻优的搜索范围,并进一步获得新冠肺炎疫情发展的精确预测模型。最后通过仿真,验证了该GABP神经网络对模型拟合预测的准确性。展开更多
风雹灾害是造成农业生产设施破坏和经济损失的主要自然灾害之一,因此有必要构建一个高效且准确的冰雹冲击力预测模型。本研究以拱形温室大棚薄膜风雹耦合试验为基础,采用粒子群优化(Particle Swarm Optimization, PSO)算法与反向传播(Ba...风雹灾害是造成农业生产设施破坏和经济损失的主要自然灾害之一,因此有必要构建一个高效且准确的冰雹冲击力预测模型。本研究以拱形温室大棚薄膜风雹耦合试验为基础,采用粒子群优化(Particle Swarm Optimization, PSO)算法与反向传播(Back Propagation, BP)神经网络相结合的方法,构建一个高效且准确的冰雹冲击力预测模型。该模型的平均绝对误差为0.22929,平均偏差误差为−0.09017,确定系数为0.99704。相较于传统线性回归预测方法,该模型可处理大数据量,适应性强,拟合效果好,且避免了传统BP模型容易陷入局部最小的缺点。Hail disasters are one of the major natural hazards causing damage to agricultural production facilities and economic losses, necessitating the development of an efficient and accurate hail impact force prediction model. This study employs a PSO-BP neural network approach, grounded in wind-hail coupling experiments on arched greenhouse films. The resultant model demonstrates superior performance with a mean absolute error (MAE) of 0.22929, a mean bias error (MBE) of −0.09017, and a determination coefficient (R2) of 0.99704. It surpasses traditional linear regression methods in handling large datasets, adaptability, fitting accuracy, and mitigating the issue of local minima in BP models.展开更多
文摘针对新冠肺炎疫情发展模型的预测问题,提出了一种基于改进遗传算法误差反向传播(genetic algorithms back propagation,GABP)神经网络算法。该算法通过对遗传算法中的交叉、变异、及适应度等函数进行了重新定义,在获得GABP神经网络最优参数的同时,提高了最优参数的泛化能力,进而缩小了GABP网络寻优的搜索范围,并进一步获得新冠肺炎疫情发展的精确预测模型。最后通过仿真,验证了该GABP神经网络对模型拟合预测的准确性。
文摘风雹灾害是造成农业生产设施破坏和经济损失的主要自然灾害之一,因此有必要构建一个高效且准确的冰雹冲击力预测模型。本研究以拱形温室大棚薄膜风雹耦合试验为基础,采用粒子群优化(Particle Swarm Optimization, PSO)算法与反向传播(Back Propagation, BP)神经网络相结合的方法,构建一个高效且准确的冰雹冲击力预测模型。该模型的平均绝对误差为0.22929,平均偏差误差为−0.09017,确定系数为0.99704。相较于传统线性回归预测方法,该模型可处理大数据量,适应性强,拟合效果好,且避免了传统BP模型容易陷入局部最小的缺点。Hail disasters are one of the major natural hazards causing damage to agricultural production facilities and economic losses, necessitating the development of an efficient and accurate hail impact force prediction model. This study employs a PSO-BP neural network approach, grounded in wind-hail coupling experiments on arched greenhouse films. The resultant model demonstrates superior performance with a mean absolute error (MAE) of 0.22929, a mean bias error (MBE) of −0.09017, and a determination coefficient (R2) of 0.99704. It surpasses traditional linear regression methods in handling large datasets, adaptability, fitting accuracy, and mitigating the issue of local minima in BP models.
