Development and Verification of Prediction Model for Grape Downy Mildew Based on Machine Learning

doi:10.13304/j.nykjdb.2022.0866

Abstract

Abstract:

Grape downy mildew is one of the major diseases on grapes， which can occur from seedling to fruit maturity， often causing destructive losses in rainy years. To accurately predict the occurrence of grape downy mildew and minimize the hazards of grape downy mildew， based on the agrometeorological data and disease occurrence data during grape growth in 2020， a prediction model for the occurrence of grape downy mildew was developed by 4 machine learning algorithms （binary logistic regression， support vector machine， decision tree and K-nearest neighbors）， and the models were verified using the data in 2021. The results showed that the decision tree model had the best evaluation indicators in the disease prediction model， which the accuracy was 94%， and the precision， recall and F₁-score was 91%， 90% and 91%， respectively. The decision tree model’s performance was still better than the other models using validation data. Therefore， the decision tree model could be further used to develop early warning systems for grape downy mildew， which should provide technical support and decision-making guidance for controlling it in production.

Key words: machine learning, grape, downy mildew, prediction model

摘要：

葡萄霜霉病是葡萄上的主要病害之一，该病从葡萄苗期到果实成熟期都可发生，多雨年份常造成毁灭性损失。为了准确预测葡萄霜霉病的发生，最大限度地降低霜霉病对葡萄的危害，基于2020年葡萄生长期间气象数据和病害发生数据，结合4种机器学习算法（二项逻辑斯蒂、支持向量机、决策树、K最近邻）构建了葡萄霜霉病发生预测模型，并用2021年数据进行验证。结果表明，决策树模型在病害发生预测模型构建中的评价指标最优，其准确率达94%，预测发生的精准率、召回率、F₁分值分别为91%、90%、91%。经验证，决策树模型对葡萄霜霉病发生的预测精度及性能均优于其他3个模型。因此，可利用此模型进一步开发葡萄霜霉病预警系统，为生产上葡萄霜霉病的防治提供技术支持和决策指导。

关键词: 机器学习, 葡萄, 霜霉病, 预测模型

CLC Number:

S663.1

Fengxia BIAN, Kaige LIU, Xinmin RONG. Development and Verification of Prediction Model for Grape Downy Mildew Based on Machine Learning[J]. Journal of Agricultural Science and Technology, 2023, 25(8): 126-137.

边凤霞, 刘凯歌, 容新民. 基于机器学习构建葡萄霜霉病预测模型及验证[J]. 中国农业科技导报, 2023, 25(8): 126-137.

Figures/Tables 15

Fig. 1 SVM model principle

Fig. 2 Decision tree model principleNote： The schematic diagram only shows the decision-making process of the model when dealing with the problem， and does not mean that the decision result of a single feature is used as the prediction result.

Fig. 3 KNN model principle

Table 1 Confusion matrix contingency table

指标

Index

样本数Number of samples

预测不发病 Prediction label： 0

预测发病Prediction label： 1

实际不发病

True label： 0

真阴性 True negative： TN

假阳性 False positive： FP

实际发病

True label： 1

假阴性 False negative：FN

真阳性 True positive：TP

Table 2 The k value consistency classification criteria

k值 k value	一致性结果 Consistent result
0≤k<0.2	极低Slight
0.2≤k<0.4	一般Fair
0.4≤k<0.6	中等Moderate
0.6≤k<0.8	高度Substantial
0.8≤k<1.0	几乎完全Almost perfect

Fig. 4 BLR model classification prediction indicatorsA： ROC curve； B： Confusion matrix

Fig. 5 SVM model classification prediction indicatorsA： ROC curve； B： Confusion matrix

Fig. 6 DT model classification prediction indicatorsA： ROC curve； B： Confusion matrix

Fig. 7 KNN model classification prediction indicatorsA： ROC curve； B： Confusion matrix

Table 3 Classification results of different prediction models

模型 Model	类别标签 Classe label	分类结果 Classification result
模型 Model	类别标签 Classe label	准确率 Accuracy/%	精准率 Precision/%	召回率 Recall/%	F₁分值 F₁-score/%
BLR	0	86	87	93	90
BLR	1	86	83	70	76
SVM	0	86	86	94	90
SVM	1	86	85	69	76
DT	0	94	95	96	95
DT	1	94	91	90	91
KNN	0	93	96	94	95
KNN	1	93	88	91	89

Fig. 8 BLR model classification prediction validation indicators in 2021A： ROC curve； B： Confusion matrix

Fig. 9 SVM model classification prediction validation indicators in 2021A： ROC curve； B： Confusion matrix

Fig. 10 DT model classification prediction validation indicators in 2021A： ROC curve； B： Confusion matrix

Fig. 11 KNN model classification prediction validation indicators in 2021A： ROC curve； B： Confusion matrix

Table 4 Classification validation results of different prediction models

模型 Model	类别标签 Classe label	分类结果 Classification result
模型 Model	类别标签 Classe label	准确率 Accuracy/%	精准率 Precision/%	召回率 Recall/%	F₁分值 F₁-score/%
BLR	0	73	73	87	80
BLR	1	73	74	53	62
SVM	0	75	76	86	81
SVM	1	75	72	59	65
DT	0	90	92	91	92
DT	1	90	87	88	88
KNN	0	84	86	87	86
KNN	1	84	80	79	80

References 39

1	芦屹,努尔孜亚·亚力麦麦提,付文君,等.新疆伊犁河谷地区葡萄霜霉病流行与气候条件的关系[J].新疆农业科学,2020,57(10):1855-1862.
	LU Y, Yalimaimaiti Nuerziya, FU W J, et al.. Research the relationship between epidemic of grape downy mildew and climatic conditions in Ili Valley of Xinjiang [J]. Xinjiang Agric. Sci., 2020, 57(10): 1855-1862.
2	吉丽丽.石河子地区葡萄霜霉病菌孢子囊时空扩散动态及品种抗病性研究[D].石河子:石河子大学,2011.
	JI L L. The dynamics of sporangium of Plasmopara viticola and resistance of different grape varieties in Shihezi region [D]. Shihezi: Shihezi Univ., 2011.
3	张玮,燕继晔,刘梅,等.葡萄霜霉病流行与预测研究进展[J].中国果树,2020,44(3):11-15.
	ZHANG W, YAN J Y, LIU M, et al.. Research progress on prediction and epidemic of grapevine downy mildew [J]. China Fruits, 2020, 44(3):11-15.
4	GILLES T. Forecasting Downy Mildew Diseases in:Advances in Downy Mildew Research—Volume Ⅱ [M]. Dordrecht: Springer, 2004:35-67.
5	ORLANDINI S, CAPPUGI A. Calibration of agrometeorological model for the simulation of grapevine downy mildew in several viticultural areas of Europe [C]// Proceedins of Fifth European Conference on Application of Meteorology. Budapest Hungary, 2001:1-7.
6	张雪雪,王斌,田洋洋,等.作物病虫害预测机理与方法研究进展[J].中国农业科技导报,2019,21(5):110-120.
	ZHANG X X, WANG B, TIAN Y Y, et al.. Research progress on forecasting mechanism and methodology for crop disease and insect pest [J]. J. Agric. Sci. Technol., 2019, 21(5):110-120.
7	GESSLER C, PERTOT I, PERAZZOLLI M. Plasmopara viticola:a review of knowledge on downy mildew of grapevine and effective disease management [J]. Phytopathol. Mediterr., 2011, 50(1):3-44.
8	THIND T S, ARORA J K, MOHAN C, et al.. Epidemiology of Powdery Mildew, Downy Mildew and Anthracnose Diseases of Grapevine [M]. Dordrecht: Kluwer Academic Publishers, 2004:621-638.
9	FRANCESCA S, SIMONA G, FRANCESCO NICOLA T, et al.. Downy mildew (Plasmopara viticola) epidemics on grapevine under climate change [J]. Global Change Biol., 2006, 12(7):1299-1307.
10	ROSSI V, CAFFI T, GIOSUE S, et al.. A mechanistic model simulating primary infections of downy mildew in grapevine [J]. Ecol. Modell., 2008, 212(3-4):480-491.
11	白玉娇,马贵龙.葡萄霜霉病流行时间动态模型及防治指标[J].吉林农业大学学报,2020,42(3):286-292.
	BAI Y J, MA G L. Dynamic model of epidemic period and control index of grape downy mildew [J]. J. Jilin Agric. Univ., 2020, 42(3):286-292.
12	于舒怡,傅俊范,刘长远,等.沈阳地区葡萄霜霉病流行时间动态及其气象影响因子分析[J].植物病理学报,2016,46(4):529-535.
	YU S Y, FU J F, LIU C Y, et al.. Epidemic temporal dynamic of grape downy mildew and its meteorological influencing factors in Shenyang, Liaoning [J]. Acta Phytopathol. Sin., 2016, 46(4):529-535.
13	BLAISE P, GESSLER C. Development of a forecast model of grape downy mildew on a microcomputer [C]// Proceedings of International Symposium on Computer Modelling in Fruit Research and Orchard Management 276, 1989:63-70.
14	WILKS D S, SHEN K W. Threshold relative humidity duration forecasts for plant disease prediction [J]. J. Appl. Meteorol. Climatol., 1991, 30(4):463-477.
15	华来庆,熊林平,孟虹,等.AR-EGARCH模型在疾病指数时间序列建模中的应用研究[J].中国卫生统计,2006,23(6):482-485.
	HUA L Q, XIONG L P, MENG H, et al.. Application of AR-EGARCH model in establishing methods of disease index time series models [J]. Chin. J. Health Stat., 2006, 23(6):482-485.
16	POUZESHIMIYAB B, FANI S R. Epidemiology and aerobiology of Pseudoperonospora cubensis in northwest Iran [J]. Ital. J. Agrometeorol., 2020, 13(2):109-116.
17	NEUFELD K N, KEINATH A P, OJIAMBO P S. Evaluation of a model for predicting the infection risk of squash and cantaloupe by Pseudoperonospora cubensis [J]. Plant Dis., 2018, 102(5):855-862.
18	LIU Y, LI D, WAN S, et al.. A long short-term memory-based model for greenhouse climate prediction [J]. Int. J. Intell. Syst., 2022, 37(1):135-151.
19	张善文,王振,王祖良.结合知识图谱与双向长短时记忆网络的小麦条锈病预测[J].农业工程学报,2020,36(12):172-178.
	ZHANG S W, WANG Z, WANG Z L. Prediction of wheat stripe rust disease by combining knowledge graph and bidirectional long short term memory network [J]. Trans. Chin. Soc. Agric. Eng., 2020, 36(12):172-178.
20	LIAKOS K G, BUSATO P, MOSHOU D, et al.. Machine learning in agriculture: a review [J]. Int. J. Sci. Technol. Res., 2018, 18(8):2674.
21	MEZEI I, LUKIC M, BERBAKOV L, et al.. Grapevine downy mildew warning system based on NB-IoT and energy harvesting technology [J]. Electronics, 2022, 11(3):356.
22	VOLPI I, GUIDOTTI D, MAMMINI M, et al.. Predicting symptoms of downy mildew, powdery mildew, and gray mold diseases of grapevine through machine learning [J]. Ital. J. Agrometeorol., 2021, 14(2):57-69.
23	CHEN M, BRUN F, RAYNAL M, et al.. Forecasting severe grape downy mildew attacks using machine learning [J/OL]. PLoS One, 2020, 15(3):e0230254 [2022-09-10]. .
24	吴宁.基于灰色关联分析和优化SVM的葡萄霜霉病短期预测[D].上海:上海海洋大学,2020.
	WU N. Short-term prediction of grape downy mildew based on grey correlation analysis and optimized SVM [J]. Shanghai: Shanghai Ocean University, 2020.
25	MENESATTI P, ANTONUCCI F, COSTA C, et al.. Multivariate forecasting model to optimize management of grape downy mildew control [J]. Vitis, 2013, 52(3):141-148.
26	FIRANJ SREMAC A, LALIC B, MARCIC M, et al.. Toward a weather-based forecasting system for fire blight and downy mildew [J/OL]. Atmosphere, 2018, 9(12):484 [2022-09-10]. .
27	边凤霞,王富霞,刘静,等.北疆葡萄设施延晚栽培病虫害防治技术[J].农业科技通讯,2019,49(9):344-345, 354.
28	王济川,郭志刚. Logistic回归模型——方法与应用[M].北京:高等教育出版社,2001:10-17.
29	JOSE C, GOYAL P, AGGRWAL P, et al.. Local deep kernel learning for efficient non-linear svm prediction [C]// Proceedings of International Conference on Machine Learning. PMLR, 2013:486-494.
30	SONG Y Y, YING L U. Decision tree methods:applications for classification and prediction [J]. Shanghai Arch. Psychiatry, 2015, 27(2):130.
31	GUO G, WANG H, BELL D, et al.. KNN model-based approach in classification [C]// Proceedings of OTM Confederated International Conferences on the move to Meaningful Internet Systems. Berlin, Heidelberg, Springer, 2003:986-996.
32	LIU K, ZHANG C, YANG X, et al.. Development of an occurrence prediction model for cucumber downy mildew in solar greenhouses based on Long Short-Term Memory neural network [J/OL]. Agronomy, 2022, 12(2):442 [2022-09-10]. .
33	宋旺.基于SVM的葡萄霜霉病病害发生预测研究[J].中小企业管理与科技,2018,11(5):104-105.
34	秦华.基于自适应神经网络的葡萄病害发生预测研究[D].保定:河北农业大学,2009.
	QIN H. Research on the prediction of grape disease based on adaptive neural networks [J]. Baoding: Hebei Agricultural University, 2009.
35	于舒怡,李柏宏,王辉,等.基于田间空气中病菌孢子囊浓度的葡萄霜霉病病情估计模型研究[J].果树学报,2021,38(10):1767-1777.
	YU S Y, LI B H, WANG H, et al.. Study on estimation model for grape downy mildew prediction based on airborne sporangium concentration of Plasmopara viticola in field [J]. J. Fruit Sci., 2021, 38(10):1767-1777.
36	RODRIGUEZ-RAJO F J, JATO V, FERNANDEZ-GONZALEZ M, et al.. The use of aerobiological methods for forecasting Botrytis spore concentrations in a vineyard [J]. Grana, 2010, 49(01):56-65.
37	CAFFI T, ROSSI V, LEGLER S E, et al.. A mechanistic model simulating ascosporic infections by Erysiphe necator, the powdery mildew fungus of grapevine [J]. Plant Pathol., 2011, 60(3):522-531.
38	GONZALEZ-DOMINGUEZ E, CAFFI T, CILIBERTI N, et al.. A mechanistic model of Botrytis cinerea on grapevines that includes weather, vine growth stage, and the main infection pathways [J/OL]. PloS ONE, 2015, 10(10):e0140444 [2022-09-10]. .
39	HILL G N, BERESFORD R M, EVANS K J. Automated analysis of aggregated datasets to identify climatic predictors of Botrytis bunch rot in wine grapes [J]. Phytopathology, 2019, 109(1):84-95.

[1]	Qianqian LU, Abuduwaili Abulimiti, Yixing HOU, Zhihui LI, Shuang WANG, Long ZHOU. Research of the Photosynthetic Characteristics of 7 Table Grape Varieties Under Compound Salt-alkali Stress [J]. Journal of Agricultural Science and Technology, 2023, 25(7): 63-76.
[2]	Shuang WANG, Yixing HOU, Linjiao FENG, Qianqian LU, Long ZHOU. Effect of Drought Stress on Anatomical Structure of Leaves in Table Grape Varieties [J]. Journal of Agricultural Science and Technology, 2023, 25(6): 40-49.
[3]	Yinan JIA, Guangdi ZHANG, Haoyu ZHANG, Chang XU, Kunming ZHANG, Jianglong WANG, Xiaojian HOU. Study on Fruit Quality of ‘Muscat Hamburg’ Grape Applied by Silicon Fertilizer in Root Zone [J]. Journal of Agricultural Science and Technology, 2023, 25(5): 215-223.
[4]	Yanfang ZHU, Yaodong BAI, Yuanyuan WANG, Yubin LI, Qilong MA, Yan HAO, Mingxin ZHAO. Effect of Shading on Fruit Shrinkage of Wine Grape ‘Syrah’ [J]. Journal of Agricultural Science and Technology, 2022, 24(1): 54-62.
[5]	BAIGETUMUER Ayitula, LI jie, CHEN Cunkun, LIU hui, LI Xiangyang, LIN Shaohua. Combined Effect of Grapefruit Seed Extract Treatment and Polyethylene Packaging on Postharvest Quality of Toona Sinensis [J]. Journal of Agricultural Science and Technology, 2021, 23(5): 116-123.
[6]	SUN Yan, HAN Bin, YUAN Junwei, LIU Changjiang, LI Minmin, YIN Yonggang, JIA Nan, GUO Zijuan, ZHAO Shengjian*. Difference Analysis of Morphology, Structure and Physiological Characteristics in Roots of Grapevines with Different Cold Resistance [J]. Journal of Agricultural Science and Technology, 2021, 23(4): 47-57.
[7]	KONG Deqian1, HE Zhenjia2*, LIU Quanzu3, WU Yongjie1. Effect of Fertilization on Yield and Quality of ‘Cabernet Sauvignon’ Grape Under Drip Irrigation at the East Foot of Helan Mountain [J]. Journal of Agricultural Science and Technology, 2021, 23(3): 148-155.
[8]	WANG Dong1, CAO Yuanbei2, JI Yaofang3, FU Yuliang3*. Effects of Different Drip Irrigation on the Growth and Quality of Wine Grapes in Hongsibu District [J]. Journal of Agricultural Science and Technology, 2021, 23(1): 154-161.
[9]	YIN Yonggang, YUAN Junwei, LIU Changjiang, HAN Bin, LI Minmin, SUN Yan, JIA Nan, GUO Zijuan, ZHAO Shengjian*. Effects of NaCl Stress on Leaf Photosynthesis and Chlorophyll Fluorescence Parameters of Vitis sp. Rootstocks [J]. Journal of Agricultural Science and Technology, 2020, 22(8): 49-55.
[10]	GAO Suhong1,2, WU Haiyan2, WEN Xiaolei2, LU Changkuan2, ZHAO Chunming2, ZHANG Qi2, DING Yuanyuan2. Spatial Distribution and Sampling Technique of the Overwintering Eggs of Apolygus lucorum in the Vineyard [J]. Journal of Agricultural Science and Technology, 2020, 22(8): 116-122.
[11]	LI Xiaowei1,WU Baoguo1*,SU Xiaohui1,CHEN Yuling1,PENG Yiqin2,YU Yonghui2, FAN Xiaohu2. Study on Estimation Model of Eucalyptus Accumulation in Guangxi Based on Decision Tree Integrated Learning [J]. Journal of Agricultural Science and Technology, 2020, 22(6): 81-90.
[12]	SHEN Yu1,2, FANG Sheng2, WANG Fengyun1, LI Zhe2, ZHANG Chen1,2, ZHENG Jiye1*. Effective Wavelengths Study on the Identification of Slight Bruises of Apples Based on Hyperspectral Imaging [J]. Journal of Agricultural Science and Technology, 2020, 22(3): 64-71.
[13]	WANG Sha, CHENG Dawei, LI Ming, GU Hong, LI Zhengyang, QI Shuai, CHEN Jinyong*. Research Progress on Seedless-Induced Production of ‘Shine Muscat’ Grape [J]. Journal of Agricultural Science and Technology, 2020, 22(2): 58-64.
[14]	TANG Zixia, LI Mengmeng*, WANG Xiaoqin, QIU Pengxun. Extraction of Grape Greenhouses From GF-2 Remote Sensing Images#br# [J]. Journal of Agricultural Science and Technology, 2020, 22(11): 95-105.
[15]	CHENG Dawei, CHEN Jinyong*, GU Hong, LU Huiran, ZHANG Yang, GUO Xizhi, QI Shuai, LI Zhengyang. Research on Floral Cluster Pruning Methods of Red Globe Grape [J]. Journal of Agricultural Science and Technology, 2019, 21(9): 44-50.