基于多粒度级联森林算法的玉米纹枯病预测  

Maize sheath blight prediction based on GC-Forest algorithm

作　　者：魏士磊 王剑雄[1] 徐玉明[2] 孙秋亚 任一帅 沈英杰 WEI Shilei;WANG Jianxiong;XU Yuming;SUN Qiuya;REN Yishuai;SHEN Yingjie(Hebei University of Architecture,Zhangjiakou,Hebei 075000;Xuanhua Vocational College of Science and Technology,Zhangjiakou,Hebei 075000)

机构地区：[1]河北建筑工程学院,河北张家口075000 [2]宣化科技职业学院,河北张家口075000

出　　处：《河北建筑工程学院学报》2024年第3期235-240,共6页Journal of Hebei Institute of Architecture and Civil Engineering

基　　金：基于人工智能技术的安防云平台研究(XY2023080)。

摘　　要：农作物安全是其实现高产的重要因素。近年来机器学习算法为玉米纹枯病预测提供了新思路,在研究玉米纹枯病以及机器学习算法的基础上,针对传统机器学习算法模型复杂且表现不佳的缺点,提出基于多粒度级联森林算法去实现玉米纹枯病的预测。应用最大最小标准化和Z-Score标准化方法对数据进行预处理,利用单变量检验和皮尔逊系数来衡量特征参数的选择,然后将选出的特征参数作为预测模型的输入变量,建立多粒度级联森林预测模型,在测试集上运行模型,用均方根误差、平均相对误差和决定系数对模型性能进行评估。研究结果表明,多粒度级联森林模型的决定系数明显高于BP神经网络和随机森林算法,具有较好的预测效果。Crop safety is an important factor to achieve high yield.In recent years,machine learning algorithms have provided new ideas for the prediction of maize leaf bacterial wilt.Based on the study of maize leaf bacterial wilt and machine learning algorithms,aiming at the shortcomings of complex model and poor performance of traditional machine learning algorithms,this paper proposed a multi-granularity cascade forest algorithm to realize the prediction of maize leaf bacterial wilt.The max-min standardization and ZScore standardization methods are used to preprocess the data,and the univariate test and Pearson coefficient are used to measure the selection of feature parameters.Then the selected feature parameters are used as the input variables of the prediction model to establish a multi-granularity cascade forest prediction model,and the model is run on the test set.The root mean square error,average relative error and determination coefficient were used to evaluate the performance of the model.The results show that the determination coefficient of multi-granularity cascade forest model is significantly higher than that of BP neural network and random forest algorithm,and it has better prediction effect.Crop safety is an important factor to achieve high yield.In recent years,machine learning algorithms have provided new ideas for the prediction of maize leaf bacterial wilt.Based on the study of maize leaf bacterial wilt and machine learning algorithms,aiming at the shortcomings of complex model and poor performance of traditional machine learning algorithms,this paper proposed a multi-granularity cascade forest algorithm to realize the prediction of maize leaf bacterial wilt.The max-min standardization and Z-Score standardization methods are used to preprocess the data,and the univariate test and Pearson coefficient are used to measure the selection of feature parameters.Then the selected feature parameters are used as the input variables of the prediction model to establish a multi-granularity cascade forest predict

关 键 词：玉米纹枯病 多粒度级联森林算法 BP神经网络 随机森林算法 

分 类 号：TP389.1[自动化与计算机技术—计算机系统结构]