基于极端学习机算法的虫情预测研究

摘要

Pest situation forecasting is of great significance to social economy, agricultural produc-tion and human life. However, the traditional modeling methods have various limitations, especially in terms of effectiveness and universality. In view of this, Extreme Learning Machine algorithm is used to re-model the pest situation forecasting problem. Through the detailed discussion of the structure and principle of ELM algorithm, two important theorems are obtained and proved, and then the learning rules of the algorithm are summed up. On this basis, the actual pest situation forecasting research work is car-ried out. By using the new modeling algorithm, MATLAB is used to simulate the actual data samples of 18 years,and the detailed prediction simulation results are obtained.The results show that the new mod-eling method using extreme learning machine algorithm has obvious advantages in effectiveness and uni-versality compared with the simulation results using other three algorithms.%虫情预测对社会经济、农业生产和人类生活有着重大的意义,然而传统的建模方法存在种种局限,尤其在有效性和通用性方面较差,鉴于此,本研究采用极端学习机算法对虫情预测问题重新建模.通过对极端学习机算法结构和原理的详细讨论,得出两条重要定理并加以证明,从而归纳出该算法的学习规则,以此为基础,进行实际的虫情预测研究工作.利用重新建模后的算法方案,使用Matlab对总计18年时间的实际数据样本进行仿真,得出详细的预测仿真结果,并与采用其他三种算法的仿真结果进行对比,结果表明采用极端学习机算法的新建模在有效性和通用性方面具有明显的优势.