粮食产量预测方法研究

发布时间：2019-05-10 22:53

【摘要】：民以食为天,粮食问题自古以来就是关系人类社会发展的重要问题。我国又是人多地少的发展中国家,经济发展给人民生活带来富裕的同时,也给我国的粮食供应造成压力。粮食供需与粮食安全问题紧密相连,粮食安全问题不仅仅是经济问题,更关系到国家的长久发展,因此需要对我国粮食产量进行有效预测,从而合理解决粮食供需平衡问题,确保国家粮食安全。本文在深入了解我国粮食现状后,着重对粮食产量的预测开展工作。首先,对时间序列分析中常用的ARIMA模型进行改进,同时分别采用改进模型和传统模型对不同时间区间的产量数据进行短期预测,结果表明,当选取较长时间区间,并且采用改进模型进行预测时,预测结果越准确。其次,提出了联合动态预测模型。对粮食产量的主要影响因子进行分析,然后根据关联度选择主要因子与产量数据共同构建模型,对选中的影响因子采用改进的ARIMA模型进行动态预测,再结合多元回归实现粮食产量的中长期动态预测。最后,考虑到产量数据的非线性特征,以及模型参数优化等问题,本文结合统计学习理论的相关内容,分析并学习了最小二乘支持向量机的应用原理和粒子群算法的全局寻优特性,提出了基于粒子群优化的最小二乘支持向量机模型,二者的有效结合使得预测模型既有更快的求解速度,又能保证参数选择全局最优,并且在数据预处理阶段加入了平滑处理,相较于未平滑处理具有更高的预测精度。实验结果表明:改进的ARIMA模型和联合动态预测模型能够有效的实现粮食产量的短期和中长期预测,并且后者的预测精度相较于前者以及传统灰色模型都有所提升。基于粒子群优化的最小二乘支持向量机模型较好的解决了非线性问题,在对粮食产量进行短期预测时预测精度更好。
[Abstract]:Since ancient times, the food problem has been an important issue related to the development of human society. China is a developing country with a large population and a small land. economic development not only brings wealth to the people's life, but also puts pressure on the food supply of our country. The problem of grain supply and demand is closely related to the problem of food security. The problem of food security is not only an economic problem, but also related to the long-term development of the country. Therefore, it is necessary to effectively predict the grain output of our country, so as to reasonably solve the problem of the balance of grain supply and demand. Ensure national food security. After deeply understanding the present situation of grain in China, this paper focuses on the prediction of grain output. First of all, the ARIMA model, which is commonly used in time series analysis, is improved, and the improved model and the traditional model are used to predict the yield data in different time intervals respectively. the results show that when the longer time interval is selected, When the improved model is used for prediction, the more accurate the prediction results are. Secondly, a joint dynamic prediction model is proposed. The main influencing factors of grain yield are analyzed, and then the model is constructed according to the correlation degree selection of the main factors and the yield data, and the selected influencing factors are dynamically predicted by the improved ARIMA model. Combined with multiple regression, the medium and long term dynamic prediction of grain yield is realized. Finally, considering the nonlinear characteristics of yield data and the optimization of model parameters, this paper combines the relevant contents of statistical learning theory. The application principle of least squares support vector machine and the global optimization characteristics of particle swarm optimization algorithm are analyzed and learned, and the least square support vector machine model based on particle swarm optimization is proposed. The effective combination of the two makes the prediction model not only solve faster, but also ensure the global optimization of parameter selection, and the smoothing processing is added in the data preprocessing stage, which has higher prediction accuracy than the unsmoothed processing. The experimental results show that the improved ARIMA model and the joint dynamic prediction model can effectively realize the short-term and medium-and long-term prediction of grain yield, and the prediction accuracy of the latter is higher than that of the former and the traditional grey model. The least square support vector machine model based on particle swarm optimization solves the nonlinear problem well, and the prediction accuracy is better when predicting grain yield in the short term.
【学位授予单位】：河南工业大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：F326.11;O211.61

【参考文献】