震灾网络舆情风险监测指标及其评估方法

为科学构建震灾网络舆情风险评估体系,在舆情监测指标构建基础上提出基于加速遗传算法的BP神经网络（AGABP）风险评估方法。依据网络舆情演化理论,围绕震灾网络舆情的物理属性和社会属性提出2个维度、4个二级指标和10个三级指标的震灾网络舆情风险监测指标体系。在评估方法上,针对常规评估系统对非线性、高维度和非正态评估问题的局限性,利用BP网络能够以任意均方差的精度逼近任意平方可积非线性连续函数的优势,将BP网络用于震灾网络舆情风险监测评估中,并用加速遗传算法（AGA）对BP网络参数进行优化,以解决常规BP网络存在训练速度慢和容易过早收敛的问题。通过随机样本数据对AGABP模型进行自学习训练,并用实际样本数据验证,结果表明：与BP神经网络、逻辑斯蒂曲线相比,本研究所构建的AGABP模型在收敛速度、评估准确度上有明显优势,能够用于震灾网络舆情风险管理实践中。

Risk monitoring indexes and assessment method of network public opinion of earthquake disaster

To construct the public opinion risk assessment system for earthquake disaster network, the paper puts forward the risk assessment method of the public opinion of Accelerating Genetic Algorithm BP neural network (AGABP) for earthquake disaster after the public opinion monitoring index is built. According to the evolution theory of network public opinion, the paper focuses on the physical and social attributes of the earthquake disaster network public opinion, and puts forward earthquake disaster Internet public opinion risk monitoring indexes of 2 dimensions, 4 indexes of second-level and 10 indexes of third-level. According to the disadvantage of the conventional evaluation system for nonlinear, high dimension and non-normal evaluation problem, this paper makes use of the advantage of BP network, which can approximate any nonlinear continuous function with arbitrary precision. The paper uses BP network for evaluating the risk of earthquake disaster network public opinion, and also uses Accelerating Genetic Algorithm (AGA) improving the shortcoming of BP to solve the problem of slow training and premature convergence in conventional BP networks. The paper uses random sample data to carry out self-learning training for AGABP model and verifies it with actual sample data. The research results show that:AGABP model has obvious advantages in convergence speed and accuracy to compare with BP neural network and logistics curve, and can be applied to risk management practice of earthquake network public opinion.