基于XGBoost的山核桃干腐病多分类预测模型研究

DOI：10.3969/j.issn.1672-3872.2023.24.002

基金项目：国家自然科学基金资助项目（050714723，U1809208）；浙江省科技厅公益项目（GN21F020001）

作者：李波¹，徐炳潮²，罗煦钦²，程云霞³

（1. 浙江农林大学数学与计算机科学学院，浙江 杭州 311300；2. 杭州市临安区农业农村信息服务中心，浙江 杭州 311300；3. 杭州市临安区太阳镇公共服务中心，浙江 杭州 311300）

摘　要：【目的】通过分析影响山核桃短期内干腐病发病程度的因素，对比不同机器学习算法，得到发病程度预测效果较好的模型，为科学、绿色防治工作提供思路。【方法】采用机器学习算法XGBoost、逻辑回归和BP神经网络作为构建预测模型的基础，以五折交叉验证方法验证模型。【结果】对山核桃干腐病发病程度的影响因素排名，由高到低分别是候平均温度、候平均湿度、病斑数目、候降水量；集成机器学习算法XGBoost构建的预测模型各评价指标都高于逻辑回归和BP神经网络；集成机器学习算法XGBoost在山核桃干腐病发病程度多分类预测问题上得到的效果优于传统机器学习算法。

关键词：山核桃干腐病；XGBoost算法；多分类预测模型

Research on Multi-Classification Prediction Model of Canker Carya Cathayensis Based on XGBoost Algorithm

Li Bo¹, Xu Bingchao², Luo Xuqin², Cheng Yunxia³

(1.School of Mathematics and Computer Science, Zhejiang Agriculture and Forestry University, Zhejiang Hangzhou 311300; 2.Hangzhou Lin’an District Agricultural and Rural Information Service Center, Zhejiang Hangzhou 311300; 3.Sun Town Public Service Center, Lin’an District, Hangzhou City, Zhejiang Hangzhou 311300)

Abstract：[Objective] By analysing the factors affecting the incidence degree of Carya cathayensis in canker in the short term, and comparing different machine learning algorithms to get the model with better prediction effect of incidence degree, we can provide ideas for scientific and green control work. [Method] The machine learning algorithms XGBoost, logistic regression and BP neural network were used as the basis for constructing the prediction model, and the model was verified by the five-fold cross-validation method. [Result] Factors affecting the incidence degree of canker Carya cathayensis disease were ranked, from high to low, as pentad average temperature, pentad average humidity, number of spots, and pentad precipitation; each evaluation index of the prediction model constructed by the integrated machine learning algorithm XGBoost was higher than that of logistic regression and BP neural network; the integrated machine learning algorithm XGBoost obtained better results than the traditional machine learning algorithm in the multi-classification prediction problem of the incidence degree of canker Carya cathayensis disease better results than traditional machine learning algorithms.

Keywords：canker Carya cathayensis; XGBoost algorithm; multi-classification prediction model

引文信息：[1]李波，徐炳潮，罗煦钦，等.基于XGBoost的山核桃干腐病多分类预测模型研究[J].南方农机，2023，54(24)：6-9.

查看全文请下载PDF文件↓