农业物联网决策系统数据处理方法研究
发布时间：2019-06-05 17:33
【摘要】：农业物联网是物联网技术在农业生产管理中的具体应用。决策系统是农业物联网中非常重要的一个部分,负责数据信息的处理、分析、推理和决策,为生产管理提供决策支持以利于农业生产。而粮食产量预测是决策系统中的重要功能模块,对农作物品种的改良、粮食生产结构的调整、国家政策方针的制定、加快我国现代化农业生产的建设和发展、保证粮食安全、国民经济发展和国家安全具有重大意义,因此粮食产量的科学预测具有重要的战略地位。本文以云南省蒙自县试验地玉米产量为例,研究农作物产量预测方法,以提高农作物产量预测精度。本文的主要内容包括以下几个方面:(1)分析国内外农作物产量预测方法的研究现状,在此基础上选择多元线性回归、人工神经网络(Artificial Neural Network,ANN)、支持向量机(Support Vector Machine,SVM)方法作为产量预测方法,并对其理论进行了分析。(2)利用多元线性回归、ANN、SVM方法对云南省蒙自县玉米产量进行预测,对各模型的预测结果、预测误差进行分析,判断各模型是否符合精度要求,并采用相关系数R、平均绝对误差MAE和均方根误差RMSE来衡量模型的总体预测能力。(3)从各模型检验结果和模型理论两个方面对建立的多元线性回归模型、ANN模型和SVM模型进行比较分析,得出最终结论。在前人的研究基础上,本论文采用多元线性回归、ANN和SVM方法,以云南省蒙自县玉米产量为例,对农作物产量预测进行了研究,得出以下主要结论:(1)所建立的多元线性回归模型、ANN模型和SVM模型均能应用于玉米产量的预测,这三个模型的预测误差都比较小,预测精度高。(2)在多元线性回归模型中,玉米产量和5月的日均参考作物腾发量、8月的最高气温、6月的最低气温、8月的最低气温关系紧密,而在ANN模型和SVM模型中,玉米产量和3月到5月的日均参考作物腾发量、8月的最高气温、6月的最低气温、8月的最低气温关系紧密。(3)针对云南省蒙自县玉米产量的预测,多元线性回归模型的R值为0.813,MAE值为0.1044,RMSE值为0.1441;ANN模型的R值为0.7479,MAE值为0.1326,RMSE值为0.1691;SVM模型的R值为0.8046,MAE值为0.1179,RMSE值为0.1546。多元线性回归模型的R值最高,MAE和RMSE值最低,因此其预测效果最好,其次是SVM模型,ANN模型最差。
[Abstract]:The Internet of Things is the specific application of Internet of Things technology in the management of agricultural production. The decision-making system is a very important part of the agricultural Internet of things, which is responsible for the processing, analysis, reasoning and decision-making of the data information, and provides decision support for the production management to facilitate the agricultural production. and the grain yield prediction is an important function module in the decision-making system, the improvement of the crop variety, the adjustment of the grain production structure, the development of the national policy guideline, the construction and the development of the modern agricultural production in China, and the food security, The development of national economy and national security are of great significance, so the scientific prediction of grain output has an important strategic position. Taking the yield of corn in Mengzi county of Yunnan province as an example, this paper studies the method of crop yield prediction to improve the accuracy of crop yield prediction. The main contents of this paper are as follows: (1) The research status of crop yield prediction method at home and abroad is analyzed. Based on this, a multi-element linear regression, an artificial neural network (ANN) and a Support Vector Machine (SVM) method are selected as the yield prediction method. The theory is analyzed. (2) using the multi-element linear regression, the ANN and the SVM method to forecast the yield of the corn in the Mengzi county of Yunnan province, and analyzing the prediction result and the prediction error of each model, judging whether each model meets the precision requirement, and adopting the correlation coefficient R; The average absolute error MAE and the root mean square error RMSE measure the overall prediction capability of the model. (3) The multi-element linear regression model, the ANN model and the SVM model are compared and analyzed from the two aspects of each model test result and the model theory to obtain the final conclusion. Based on the previous research, this paper uses the method of multiple linear regression, ANN and SVM to study the crop yield prediction by taking the maize yield of Mengzi County in Yunnan province as an example. The main conclusions are as follows: (1) the established multi-component linear regression model, Both the ANN model and the SVM model can be applied to the prediction of corn yield, and the prediction error of the three models is small and the prediction accuracy is high. (2) In the multiple linear regression model, the corn yield and the average daily reference crop evapotranspiration in May, the highest air temperature in August, the minimum temperature in June and the lowest temperature in August, and in the ANN model and the SVM model, The corn yield and the average daily reference crop evapotranspiration from March to May, the highest temperature in August, the minimum temperature in June, and the lowest temperature in August. (3) The R value of the multiple linear regression model is 0.813, the MAE value is 0.1044, the RMSE value is 0.1441, the R value of the ANN model is 0.7479, the MAE value is 0.1326, the RMSE value is 0.1691, the R value of the SVM model is 0.8046, the MAE value is 0.1179, and the RMSE value is 0.1546. The R value of the multiple linear regression model is the highest, and the value of MAE and RMSE is the lowest, so the prediction effect is the best, second is the SVM model, and the ANN model is the worst.
【学位授予单位】：电子科技大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：TP391.44;TN929.5
本文编号：2493702
[Abstract]:The Internet of Things is the specific application of Internet of Things technology in the management of agricultural production. The decision-making system is a very important part of the agricultural Internet of things, which is responsible for the processing, analysis, reasoning and decision-making of the data information, and provides decision support for the production management to facilitate the agricultural production. and the grain yield prediction is an important function module in the decision-making system, the improvement of the crop variety, the adjustment of the grain production structure, the development of the national policy guideline, the construction and the development of the modern agricultural production in China, and the food security, The development of national economy and national security are of great significance, so the scientific prediction of grain output has an important strategic position. Taking the yield of corn in Mengzi county of Yunnan province as an example, this paper studies the method of crop yield prediction to improve the accuracy of crop yield prediction. The main contents of this paper are as follows: (1) The research status of crop yield prediction method at home and abroad is analyzed. Based on this, a multi-element linear regression, an artificial neural network (ANN) and a Support Vector Machine (SVM) method are selected as the yield prediction method. The theory is analyzed. (2) using the multi-element linear regression, the ANN and the SVM method to forecast the yield of the corn in the Mengzi county of Yunnan province, and analyzing the prediction result and the prediction error of each model, judging whether each model meets the precision requirement, and adopting the correlation coefficient R; The average absolute error MAE and the root mean square error RMSE measure the overall prediction capability of the model. (3) The multi-element linear regression model, the ANN model and the SVM model are compared and analyzed from the two aspects of each model test result and the model theory to obtain the final conclusion. Based on the previous research, this paper uses the method of multiple linear regression, ANN and SVM to study the crop yield prediction by taking the maize yield of Mengzi County in Yunnan province as an example. The main conclusions are as follows: (1) the established multi-component linear regression model, Both the ANN model and the SVM model can be applied to the prediction of corn yield, and the prediction error of the three models is small and the prediction accuracy is high. (2) In the multiple linear regression model, the corn yield and the average daily reference crop evapotranspiration in May, the highest air temperature in August, the minimum temperature in June and the lowest temperature in August, and in the ANN model and the SVM model, The corn yield and the average daily reference crop evapotranspiration from March to May, the highest temperature in August, the minimum temperature in June, and the lowest temperature in August. (3) The R value of the multiple linear regression model is 0.813, the MAE value is 0.1044, the RMSE value is 0.1441, the R value of the ANN model is 0.7479, the MAE value is 0.1326, the RMSE value is 0.1691, the R value of the SVM model is 0.8046, the MAE value is 0.1179, and the RMSE value is 0.1546. The R value of the multiple linear regression model is the highest, and the value of MAE and RMSE is the lowest, so the prediction effect is the best, second is the SVM model, and the ANN model is the worst.
【学位授予单位】：电子科技大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：TP391.44;TN929.5
【参考文献】
相关期刊论文 前1条
1 吴玉鸣,李建霞;通径分析在我国粮食生产相关研究中的应用[J];广西师范大学学报(自然科学版);2003年03期
，本文编号：2493702
本文链接：https://www.wllwen.com/kejilunwen/wltx/2493702.html
