[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"item-3280":3,"related-3280":66},{"id":4,"title":5,"url":6,"summary":7,"summary_zh":8,"content":9,"source_name":10,"source_url":6,"published_at":11,"category":12,"cover_url":9,"hotness":13,"is_selected":14,"score":15,"score_detail":16,"sources":23,"tags":25,"search_phrases":31,"slug":34,"view_count":35,"doi":36,"paper":37,"created_at":65},3280,"Machine Learning for Surface Water Quality Management in Malaysia: Integrating Physicochemical Parameters, Nutrient Loads, Land Use, Hydrological Conditions, and Anthropogenic Pressure","https:\u002F\u002Fdoi.org\u002F10.64440\u002Fijes\u002Fengineeringx0021","Surface water quality in Malaysia is increasingly influenced by rapid urbanization, industrial expansion, agricultural activities, wastewater discharge, climatic variability, and changes in hydrological regimes. These interacting pressures create complex and nonlinear relationships among physicochemical water-quality parameters, watershed characteristics, and pollutant dynamics, making conventional monitoring approaches insufficient for comprehensive and timely assessment. This study develops a data-driven machine learning framework for surface water quality assessment, prediction, pollution-event detection, and management in Malaysian river systems. The proposed framework integrates physicochemical, hydrological, spatial, land-use, and anthropogenic variables to characterize the temporal and spatial variability of water-quality conditions. The investigated parameters include dissolved oxygen, biochemical oxygen demand, chemical oxygen demand, pH, temperature, electrical conductivity, turbidity, total suspended solids, ammoniacal nitrogen, nitrate, phosphate, total nitrogen, and total phosphorus, together with rainfall, streamflow, water level, antecedent rainfall, urbanization, agricultural land use, industrial land use, population density, wastewater pressure, and road density. Seven machine learning approaches are incorporated, including Decision Trees (DT), Random Forest (RF), Support Vector Machines (SVM), Support Vector Regression (SVR), Artificial Neural Networks (ANN), Long Short-Term Memory (LSTM) networks, and hybrid ensemble models. The Malaysian National Water Quality Standards and Water Quality Index (WQI) are incorporated to establish the regulatory and quantitative basis for water-quality classification and prediction. The framework further integrates explainable artificial intelligence techniques, including SHAP and LIME, to quantify predictor contributions and identify the dominant environmental and anthropogenic factors governing model outputs. The proposed methodology enables simultaneous water-quality classification, continuous prediction, anomaly and pollution-event detection, source-pressure identification, and management-oriented decision support. Particular emphasis is placed on temporal variability, nonlinear interactions, hydrological influence, land-use effects, and the integration of multiple environmental data sources. The study provides a comprehensive Malaysia-oriented machine learning framework that can enhance conventional surface-water monitoring by combining predictive modeling, explainability, and integrated watershed information. The findings establish a methodological basis for developing real-time and transferable water-quality prediction systems capable of supporting pollution mitigation, watershed management, and sustainable water-resource planning across Malaysian river basins.","马来西亚地表水水质日益受到快速城市化、工业扩张、农业活动、废水排放、气候变率以及水文情势变化的影响。这些相互作用压力在物理化学水质参数、流域特征与污染物动态之间形成了复杂的非线性关系，使传统监测方法难以实现全面且及时的评估。本研究开发了一个数据驱动的机器学习框架，用于马来西亚河流系统的地表水水质评估、预测、污染事件检测与管理。所提出的框架整合了物理化学、水文、空间、土地利用和人为变量，以刻画水质条件的时间与空间变异性。所研究的参数包括溶解氧、生化需氧量、化学需氧量、pH、温度、电导率、浊度、总悬浮固体、氨氮、硝酸盐、磷酸盐、总氮和总磷，以及降雨量、河流流量、水位、前期降雨、城市化、农业用地、工业用地、人口密度、废水压力和道路密度。该框架纳入了七种机器学习方法，包括决策树（DT）、随机森林（RF）、支持向量机（SVM）、支持向量回归（SVR）、人工神经网络（ANN）、长短期记忆（LSTM）网络以及混合集成模型。研究纳入马来西亚国家水质标准和水质指数（WQI），以建立水质分类与预测的监管和定量基础。该框架进一步整合了可解释人工智能技术，包括SHAP和LIME，以量化预测因子的贡献并识别主导模型输出的主要环境与人为因素。所提出的方法能够同时实现水质分类、连续预测、异常与污染事件检测、源压力识别以及面向管理的决策支持。研究特别强调时间变异性、非线性相互作用、水文影响、土地利用效应以及多环境数据源的整合。该研究提供了一个面向马来西亚的综合机器学习框架，可增强传统",null,"The International Journal of Engineering Sciences","2026-09-22T00:00:00Z","论文",10,false,68,{"impact":17,"substance":18,"depth":19,"authority":13,"freshness":20,"relevant":21,"comment":22},12,20,17,9,1,"马来西亚河流水质机器学习框架，方法体系完整但属区域性研究，对国内农业信息化参考价值有限。",[24],{"name":10,"url":6},[26,27,28,29,30],"农业人工智能","机器学习","可解释AI","水质监测","流域管理",[32,33],"马来西亚 地表水 机器学习","水质指数 机器学习 预测","马来西亚地表水机器学习-3280",0,"10.64440\u002Fijes\u002Fengineeringx0021",{"doi":36,"openalex_id":38,"authors":39,"venue":10,"cited_by_count":35,"oa_url":6,"card":58,"direction":62,"ingested_from":64},"W7214031821",[40,42,44,46,48,50,52,54,56],{"name":41,"orcid":9},"Iskandar M. Wahyu",{"name":43,"orcid":9},"Gusti E. Rosyadi",{"name":45,"orcid":9},"May Abdul Hafed Abdul kader",{"name":47,"orcid":9},"Nuryani U. Humaira",{"name":49,"orcid":9},"Safaruddin P. Nasyita",{"name":51,"orcid":9},"Sudarijati Sudarijati;",{"name":53,"orcid":9},"Ujang Asmil Zuwariah",{"name":55,"orcid":9},"Anisa Ricardi",{"name":57,"orcid":9},"Achuo Azmaine",{"tldr":59,"method":60,"finding":61,"direction":62,"opportunity":63},"构建马来西亚地表水质机器学习框架，融合多源环境数据实现分类、预测与污染事件检测。","用DT、RF、SVM、SVR、ANN、LSTM及混合集成模型，结合SHAP\u002FLI","多源变量与可解释AI能有效刻画水质非线性时空变化并识别主要人为压力源。","农业人工智能与决策模型","可迁移至中国流域，探索农业面源污染与水文气候耦合下的可解释水质预警模型。","openalex","2026-09-23T23:30:32.639861Z",{"total":67,"page":21,"page_size":67,"items":68},6,[69,119,162,205,232,266],{"id":70,"title":71,"url":72,"summary":73,"summary_zh":74,"content":9,"source_name":75,"source_url":72,"published_at":76,"category":12,"cover_url":9,"hotness":13,"is_selected":14,"score":77,"score_detail":78,"sources":83,"tags":85,"search_phrases":88,"slug":91,"view_count":35,"doi":92,"paper":93,"created_at":118},3076,"Research on the Inversion Method of Total Phosphorus Concentration in Water Bodies Based on PSO-Adam-BP Neural Network Model","https:\u002F\u002Fdoi.org\u002F10.15244\u002Fpjoes\u002F220960","This study investigated Baiyangdian Lake in the Xiong’an New Area to invert total phosphorus (TP) concentrations using Sentinel-2 multispectral imagery and in situ measurements, aiming to support water-quality assessment and eutrophication management. Based on the multispectral remote sensing data of Sentinel-2 and measured total phosphorus data of water bodies, the study used specific feature combinations such as NDTI, B5, B11-B12 difference, and sensitive combination bands of measured data as model inputs to construct a PSO-Adam-BP neural network machine learning model for total phosphorus concentration inversion. Compared with the traditional BP and PSO-BP baseline models, the proposed framework significantly improved the accuracy of the model, with an R2 value reaching 0.9351. Moreover, it successfully reduced the average relative error by up to 53.5%, from 4.80% to 2.24%, and decreased the maximum relative error by up to 31.7%, from 11.21% to 7.66%, demonstrating the model’s highly robust and precise inversion performance. This study provides a new method for water quality detection in Baiyangdian Lake and is of great significance to the protection and development of the water environment in Xiong’an New Area.","本研究以雄安新区白洋淀为研究对象，利用Sentinel-2多光谱影像和实地测量数据反演总磷（TP）浓度，旨在为水质评价和富营养化管理提供支持。基于Sentinel-2多光谱遥感数据和白洋淀水体总磷实测数据，研究采用NDTI、B5、B11-B12差值等特征组合及实测数据的敏感组合波段作为模型输入，构建了PSO-Adam-BP神经网络机器学习模型用于总磷浓度反演。与传统BP和PSO-BP基准模型相比，所提出的框架显著提高了模型精度，R²值达到0.9351。此外，该模型成功将平均相对误差最多降低53.5%，从4.80%降至2.24%，并将最大相对误差最多降低31.7%，从11.21%降至7.66%，展现出高度稳健且精确的反演性能。本研究为白洋淀水质检测提供了新方法，对雄安新区水环境保护与开发具有重要意义。","Polish Journal of Environmental Studies","2026-09-18T00:00:00Z",72,{"impact":79,"substance":80,"depth":19,"authority":81,"freshness":67,"relevant":21,"comment":82},15,21,13,"基于Sentinel-2与PSO-Adam-BP模型实现白洋淀总磷浓度高精度反演，方法新颖、数据可靠，对雄安水环境智慧监测有实用价值。",[84],{"name":75,"url":72},[26,27,86,29,87],"遥感","白洋淀",[89,90],"白洋淀 总磷 遥感反演","Sentinel-2 水质 反演","白洋淀总磷遥感反演-3076","10.15244\u002Fpjoes\u002F220960",{"doi":92,"openalex_id":94,"authors":95,"venue":75,"cited_by_count":35,"oa_url":111,"card":112,"direction":116,"ingested_from":64},"W7213537374",[96,98,100,103,106,108],{"name":97,"orcid":9},"Guanxing Wang",{"name":99,"orcid":9},"Cui Jia",{"name":101,"orcid":102},"Linghan Gao","https:\u002F\u002Forcid.org\u002F0000-0002-8537-8787",{"name":104,"orcid":105},"Shan An","https:\u002F\u002Forcid.org\u002F0000-0001-7796-6952",{"name":107,"orcid":9},"Jia Xi",{"name":109,"orcid":110},"Yaxue Liu","https:\u002F\u002Forcid.org\u002F0009-0006-9025-8030","https:\u002F\u002Fwww.pjoes.com\u002Fpdf-220960-145347?filename=Research-on-the-Inversion.pdf",{"tldr":113,"method":114,"finding":115,"direction":116,"opportunity":117},"用PSO-Adam-BP神经网络结合Sentinel-2影像反演白洋淀总磷浓度。","Sentinel-2多光谱与实测数据，NDTI等特征组合，PSO-Adam-BP","R²达0.9351，平均相对误差由4.80%降至2.24%，精度显著优于BP与PSO-BP。","农业遥感与作物表型","可迁移至其他内陆水体与多参数水质反演，探索时序泛化与跨区域迁移能力。","2026-09-21T23:30:25.856035Z",{"id":120,"title":121,"url":122,"summary":123,"summary_zh":124,"content":9,"source_name":125,"source_url":122,"published_at":126,"category":12,"cover_url":9,"hotness":13,"is_selected":14,"score":127,"score_detail":128,"sources":133,"tags":135,"search_phrases":137,"slug":140,"view_count":35,"doi":141,"paper":142,"created_at":161},2433,"Hierarchical spectral ensemble with physics-informed augmentation for global water quality retrieval from hyperspectral remote sensing","https:\u002F\u002Fdoi.org\u002F10.3389\u002Ffrsen.2026.1860176","Retrieving water quality variables from remotely sensed reflectance spectra (Rrs) with global-scale, cross-waterbody applicability remains a grand challenge in environmental remote sensing. Here we introduce Hierarchical Spectral Ensemble (HSE), a machine learning pipeline for concurrent retrieval of four ecologically relevant water quality parameters (chlorophyll-a (Chl-a), total suspended solids (TSS), coloured dissolved organic matter absorption coefficient at 440 nm (aCDOM(440)), and Secchi depth (Zsd)), applied to the GLORIA 2022 global dataset of 7,572 co-located hyperspectral in situ ground truth measurements spanning six continents. HSE leverages a combination of five diverse base learners trained with 10-fold out-of-fold stacking, and integrated with a diversity-constrained non-negative least squares (NNLS) meta-learner, supported by a 291-dimensional feature suite spanning spectral, spatial, and temporal representations. We propose Interpolation-based Spectral Data Augmentation (ISDA), an ecologically inspired, class-balance oversampling methodology applied to the training set after the initial stratified split to prevent information leakage. Evaluated on the held-out global test set, HSE attains R2 = 0.822, 0.704, 0.841, and 0.962 for Chl-a, TSS, aCDOM(440), and Zsd respectively (mean R2 = 0.832; all in original physical units following back-transformation). Stratified analysis on a per-quartile basis reveals negative values of R2 in low-concentration regimes for Chl-a, TSS, and aCDOM(440), illustrating how global metrics can mask failures in retrieval across a majority of the distribution, specifically in the oligotrophic regime. This constitutes the primary known limitation of the method: reliable prediction in low-concentration, low-optical-signal regimes remains elusive and is demonstrated only in high-concentration, eutrophic samples where a measurable optical signal exists.","从遥感反射光谱（Rrs）中反演水质变量并实现全球尺度、跨水体的适用性，仍是环境遥感领域的一项重大挑战。本文提出分层光谱集成（Hierarchical Spectral Ensemble, HSE），一种机器学习流程，用于同时反演四个与生态相关的水质参数（叶绿素a（Chl-a）、总悬浮固体（TSS）、440 nm处有色溶解有机物吸收系数（aCDOM(440)）和透明度（Zsd）），并应用于GLORIA 2022全球数据集，该数据集包含横跨六大洲的7,572组同步高光谱原位地面实测数据。HSE结合了五种不同的基学习器，采用10折折外堆叠进行训练，并与多样性约束的非负最小二乘（NNLS）元学习器集成，辅以涵盖光谱、空间和时间表示的291维特征集。我们提出基于插值的光谱数据增强（Interpolation-based Spectral Data Augmentation, ISDA），这是一种受生态学启发的类平衡过采样方法，在初始分层划分后应用于训练集以防止信息泄漏。在留出的全球测试集上评估，HSE在Chl-a、TSS、aCDOM(440)和Zsd上分别达到R2 = 0.822、0.704、0.841和0.962（平均R2 = 0.832；均在反变换后以原始物理单位表示）。基于四分位数的分层分析揭示了Chl-a、TSS和aCDOM(440)在低浓度条件下R2为负值，说明全球指标可能掩盖分布中大多数区域的检索失败，特别是在贫营养条件下。这构成了该方法已知的主要局限性：在低浓度、低光学信号条件下实现可靠预测仍然难以实现，仅在存在可测量光学信号的高浓度、富营养化样本中得到验证。","Frontiers in Remote Sensing","2026-09-11T00:00:00Z",79,{"impact":129,"substance":130,"depth":129,"authority":81,"freshness":131,"relevant":21,"comment":132},18,22,8,"基于GLORIA全球高光谱数据集提出分层光谱集成方法，可同步反演叶绿素a、悬浮物等四项水质参数，方法新颖、数据规模大，但低浓度水体反演失效的局限也交代清楚，对农业水环境遥感监测有参考价值。",[134],{"name":125,"url":122},[26,27,86,29,136],"水环境",[138,139],"农业人工智能 机器学习 水质监测 水环境","农业人工智能 机器学习","农业人工智能机器学习水质监测水环境-2433","10.3389\u002Ffrsen.2026.1860176",{"doi":141,"openalex_id":143,"authors":144,"venue":125,"cited_by_count":35,"oa_url":155,"card":156,"direction":116,"ingested_from":64},"W7212323204",[145,147,149,151,153],{"name":146,"orcid":9},"Amirthalakshmi TM",{"name":148,"orcid":9},"Hemanth S",{"name":150,"orcid":9},"Ganesan PV",{"name":152,"orcid":9},"Rahul SG",{"name":154,"orcid":9},"Gayathri M","https:\u002F\u002Fwww.frontiersin.org\u002Fjournals\u002Fremote-sensing\u002Farticles\u002F10.3389\u002Ffrsen.2026.1860176\u002Fpdf",{"tldr":157,"method":158,"finding":159,"direction":116,"opportunity":160},"提出分层光谱集成模型HSE，从高光谱遥感反射率中全球尺度反演四种水质参数。","用GLORIA 2022全球7572条高光谱数据，结合10折堆叠集成、NNLS元","整体R²达0.832，但低浓度贫营养水体中Chl-a、TSS、aCDOM(440)的R²为负，反演失","低浓度、弱光学信号水体的反演仍是空白，可探索物理约束与少样本学习提升贫营养水体精度。","2026-09-14T23:30:27.361699Z",{"id":163,"title":164,"url":165,"summary":166,"summary_zh":167,"content":9,"source_name":168,"source_url":165,"published_at":169,"category":12,"cover_url":9,"hotness":13,"is_selected":14,"score":127,"score_detail":170,"sources":174,"tags":176,"search_phrases":180,"slug":183,"view_count":35,"doi":184,"paper":185,"created_at":204},3368,"A PCA-based deep feature optimization framework for explainable orange fruit disease classification","https:\u002F\u002Fdoi.org\u002F10.1186\u002Fs12870-026-09984-8","Accurate classification of orange fruit diseases is important for precision agriculture and yield protection. This study develops and rigorously benchmarks a hybrid deep-feature framework for classifying Black Spot, Canker, Fresh, and Greening oranges (1,090 images), combining deep feature extraction, PCA-based dimensionality reduction, and classical machine-learning classification. Eight backbones (seven CNNs and a Vision Transformer, ViT) and four classifiers (32 configurations in total) were evaluated under 5 × 5 repeated stratified cross-validation, with PCA fitted exclusively on training-fold features in every iteration to eliminate data leakage. The proposed ViT + PCA+SVM configuration achieved the highest mean accuracy, 99.12%±0.71%, significantly outperforming every CNN-based backbone, including DenseNet201 + PCA + SVM (98.48%±0.81%, p \u003C 0.001). A dedicated variance-retention sensitivity analysis justifies the 98% threshold used throughout, and ablation experiments confirm that PCA substantially reduces feature dimensionality (by ~ 55.7% for ViT and ~ 76.6% for DenseNet201) without a significant loss of accuracy for either backbone. Explainability analysis — occlusion sensitivity and SHAP for the proposed ViT model, and Grad-CAM and SHAP for the DenseNet201 comparison model — shows that both configurations base predictions on biologically relevant, disease-affected regions of the fruit rather than spurious cues. These results identify ViT + PCA+SVM as the most accurate configuration evaluated, with DenseNet201 + PCA + SVM as a closely competitive, more compact convolutional alternative for intelligent orchard disease-monitoring systems.","橙类果实病害的准确分类对精准农业和产量保护具有重要意义。本研究开发并严格基准测试了一种混合深度特征框架，用于对黑斑病、溃疡病、新鲜和黄龙病橙类（1，090张图像）进行分类，该框架结合了深度特征提取、基于PCA的降维和经典机器学习分类。在5×5重复分层交叉验证下评估了八种骨干网络（七种CNN和一种视觉Transformer，ViT）和四种分类器（共32种配置），每次迭代中PCA仅在训练折特征上拟合以消除数据泄漏。所提出的ViT + PCA+SVM配置取得了最高平均准确率，为99.12%±0.71%，显著优于所有基于CNN的骨干网络，包括DenseNet201 + PCA + SVM（98.48%±0.81%，p \u003C 0.001）。专门的方差保留敏感性分析证明了全程使用的98%阈值是合理的，消融实验证实PCA大幅降低了特征维度（ViT约降低55.7%，DenseNet201约降低76.6%），且两种骨干网络均无显著准确率损失。可解释性分析——对所提出的ViT模型采用遮挡敏感性和SHAP，对DenseNet201对比模型采用Grad-CAM和SHAP——表明两种配置均基于果实中生物学相关的病害影响区域而非虚假线索进行预测。这些结果确定ViT + PCA+SVM为所评估的最准确配置，而DenseNet201 + PCA + SVM则是一种竞争力接近且更紧凑的卷积替代方案，可用于智能果园病害监测系统。","BMC Plant Biology","2026-09-23T00:00:00Z",{"impact":171,"substance":130,"depth":129,"authority":172,"freshness":20,"relevant":21,"comment":173},16,14,"方法严谨、数据与消融实验充分，对果园智能病害监测有实用参考价值，但属细分技术论文，产业影响有限。",[175],{"name":168,"url":165},[177,26,28,178,179],"智慧农业","病害识别","柑橘种植",[181,182],"柑橘病害 深度学习 分类","ViT PCA SVM 病害识别","柑橘病害深度学习分类-3368","10.1186\u002Fs12870-026-09984-8",{"doi":184,"openalex_id":186,"authors":187,"venue":168,"cited_by_count":35,"oa_url":165,"card":199,"direction":62,"ingested_from":64},"W7214068709",[188,190,192,194,197],{"name":189,"orcid":9},"Amruta Hingmire",{"name":191,"orcid":9},"Avinash Golande",{"name":193,"orcid":9},"Vinodkumar Bhutnal",{"name":195,"orcid":196},"Sagar Dhanraj Pande","https:\u002F\u002Forcid.org\u002F0000-0003-4506-6997",{"name":198,"orcid":9},"Tanuja Pande",{"tldr":200,"method":201,"finding":202,"direction":62,"opportunity":203},"提出PCA深度特征优化框架，用ViT+SVM分类橙子病害，准确率达99.12%。","8种骨干网络提取特征，PCA降维，4种分类器，5×5交叉验证。","ViT+PCA+SVM最优，PCA降维超55%且精度不降，可解释性验证有效。","可探索轻量化模型在移动端或边缘设备的实时病害检测与多作物泛化。","2026-09-24T23:30:34.116938Z",{"id":206,"title":207,"url":208,"summary":209,"summary_zh":9,"content":9,"source_name":210,"source_url":208,"published_at":11,"category":12,"cover_url":9,"hotness":13,"is_selected":14,"score":15,"score_detail":211,"sources":213,"tags":215,"search_phrases":218,"slug":221,"view_count":35,"doi":222,"paper":223,"created_at":231},3274,"Farmer-Friendly Decision Support System Using Explainable Orangutan Artificial Intelligence for Soil Fertility","https:\u002F\u002Fdoi.org\u002F10.1007\u002Fs42979-026-05329-2","Farmer-Friendly Decision Support System Using Explainable Orangutan Artificial Intelligence for Soil Fertility。SN Computer Science","SN Computer Science",{"impact":17,"substance":129,"depth":171,"authority":81,"freshness":20,"relevant":21,"comment":212},"论文提出可解释猩猩优化算法驱动的土壤肥力决策支持系统，方法新颖且面向农户，但尚属学术探索阶段，产业影响有限。",[214],{"name":210,"url":208},[177,26,28,216,217],"决策支持系统","土壤肥力",[219,220],"农业人工智能 决策支持系统 土壤肥力 智慧农业","农业人工智能 决策支持系统","农业人工智能决策支持系统土壤肥力智慧农业-3274","10.1007\u002Fs42979-026-05329-2",{"doi":222,"openalex_id":224,"authors":225,"venue":210,"cited_by_count":35,"oa_url":9,"card":9,"direction":230,"ingested_from":64},"W7214018467",[226,228],{"name":227,"orcid":9},"K. Komala Devi",{"name":229,"orcid":9},"Josephine Prem Kumar","智慧农业 \u002F 农业物联网","2026-09-23T23:30:12.314548Z",{"id":233,"title":234,"url":235,"summary":236,"summary_zh":237,"content":9,"source_name":238,"source_url":235,"published_at":11,"category":12,"cover_url":9,"hotness":239,"is_selected":14,"score":240,"score_detail":241,"sources":243,"tags":247,"search_phrases":250,"slug":253,"view_count":35,"doi":254,"paper":255,"created_at":265},3271,"A Systematic Study of Supervised and Ensemble Learning Approaches for Crop Selection in Smart Agriculture","https:\u002F\u002Fdoi.org\u002F10.5281\u002Fzenodo.22896343","Selecting the crop best matched to a field’s soil and climate is one of the highest-leverage decisions in agriculture, and one that farmers have traditionally made by intuition or inherited practice. Soil pH together with nitrogen, phosphorus and potassium concentration, and the local temperature, humidity and rainfall regime, jointly determine which crop will flourish and which will fail, and the relationship between those variables and crop performance is neither linear nor independent. Machine learning has therefore become the dominant approach to automated crop recommendation. This paper reviews the field across four technique families — classical supervised learning, ensemble and boosting methods, deep learning and metaheuristic hybrids, and IoT and deployment-oriented systems — and compares twenty-four representative studies published between 2016 and 2026 in terms of method, data source, reported accuracy, advantage and limitation. A generic seven-stage recommendation pipeline is presented and each family is situated within it. The comparison shows that reported accuracy on the standard nutrient-and-climate benchmark has converged in a narrow band between roughly 98 and 99.5 per cent, that boosting and ensemble methods occupy the upper part of that band, and that further gains on the benchmark are no longer the binding constraint on the field. The gaps that remain open are instead the absence of socio-economic and market variables from the decision, the lack of region-specific and long-horizon environmental validation, dataset narrowness and geographic bias, limited interpretability, and the accessibility of these systems to small and resource-poor farmers. These are consolidated into a set of research directions for future work.","选择与田块土壤和气候最匹配的作物是农业中杠杆效应最高的决策之一，而农民传统上依靠直觉或世代相传的经验来做出这一决策。土壤pH值以及氮、磷、钾浓度，加上当地的气温、湿度和降雨状况，共同决定了哪种作物能够茁壮成长、哪种会歉收，而这些变量与作物表现之间的关系既非线性也非相互独立。因此，机器学习已成为自动化作物推荐的主流方法。本文从四个技术族系——经典监督学习、集成与提升方法、深度学习与元启发式混合方法，以及物联网与面向部署的系统——对该领域进行了综述，并从方法、数据来源、报告精度、优势和局限性方面比较了2016年至2026年间发表的二十四项代表性研究。本文提出了一个通用的七阶段推荐流程，并将每个技术族系置于该流程中加以定位。比较结果表明，在标准养分与气候基准上的报告精度已收敛于约98%至99.5%的狭窄区间内，提升与集成方法占据该区间的上端，而在该基准上进一步提升已不再是该领域的约束瓶颈。真正尚未填补的空白在于：决策中缺乏社会经济和市场变量，缺少针对特定区域和长期环境验证，数据集狭窄且存在地理偏差，可解释性有限，以及这些系统对小型和资源匮乏农户的可及性不足。这些空白被归纳为未来工作的一系列研究方向。","Zenodo (CERN European Organization for Nuclear Research)",25,78,{"impact":171,"substance":130,"depth":129,"authority":81,"freshness":20,"relevant":21,"comment":242},"系统综述24项研究并指出基准精度已趋饱和，真正瓶颈转向社会经济变量与可解释性，对智慧农业选种方向有参考价值。",[244,245],{"name":238,"url":235},{"name":238,"url":246},"https:\u002F\u002Fdoi.org\u002F10.5281\u002Fzenodo.22896344",[177,26,27,248,249],"精准农业","作物推荐",[251,252],"农业人工智能 作物推荐 智慧农业 机器学习","农业人工智能 作物推荐","农业人工智能作物推荐智慧农业机器学习-3271","10.5281\u002Fzenodo.22896343",{"doi":254,"openalex_id":256,"authors":257,"venue":238,"cited_by_count":35,"oa_url":235,"card":260,"direction":230,"ingested_from":64},"W7214002748",[258],{"name":259,"orcid":9},"Prof. Nagendra Patel Sahil Verma",{"tldr":261,"method":262,"finding":263,"direction":62,"opportunity":264},"系统综述2016-2026年24项作物推荐研究，比较四类机器学习方法并指出基准精度已趋饱和。","综述监督学习、集成提升、深度学习与元启发式、物联网部署四类方法及七阶段流程。","标准基准精度收敛于98%-99.5%，提升集成法最优，但精度已非领域瓶颈。","将社会经济与市场变量、区域长期环境验证及可解释性纳入作物推荐，服务小农户。","2026-09-23T23:30:09.369987Z",{"id":267,"title":268,"url":269,"summary":270,"summary_zh":271,"content":9,"source_name":272,"source_url":269,"published_at":273,"category":12,"cover_url":9,"hotness":13,"is_selected":14,"score":274,"score_detail":275,"sources":277,"tags":279,"search_phrases":282,"slug":285,"view_count":35,"doi":286,"paper":287,"created_at":323},3258,"Machine learning models combined with feature importance methods for honey yield classification: A replicable approach","https:\u002F\u002Fdoi.org\u002F10.1016\u002Fj.atech.2026.102575","Beekeepers require planning tools supported by modern technologies, such as machine learning and the Internet of Things, to address agricultural challenges such as the decrease and irregularity in honey production. To ensure replicability, this article presents a research workflow that begins with the creation of an open-access database, developed from annual production records and climatic variables (temperature and rainfall), integrating data construction, explainability analysis, and model evaluation. Then, feature importance methods and explainability techniques are applied, such as feature importance, the depth-wise frequency of each feature in random forest, and the Shapley Additive Explanations method. Finally, machine learning approaches are evaluated for honey yield prediction: logistic regression, k-nearest neighbors, support vector machine, decision tree, multilayer perceptron, random forest, linear discriminant analysis, gradient boosting, and Naive Bayes. These algorithms are compared considering: (1) a baseline corresponding to models without hyperparameter optimization, using leave-one-out cross-validation and stratified 10-fold cross-validation; (2) the baseline plus normalization\u002Fstandardization (div-max, min-max, and z-score); (3) the configuration in point 2 plus bagging; (4) evaluation of a data-augmentation and class-balancing strategy using SMOTE, together with model combination via the Voting Classifier. The results suggest that rainfall is one of the most important variables for honey yield prediction. By selecting certain features, the models improve in some cases or do not significantly degrade their performance. The min-max and z-score methods led to improved predictions in some algorithms; for example, support vector machine achieved an accuracy of 0.80, compared with the 0.67 accuracy reported in the reference study based on random forest, representing an increase of 13 percentage points. Finally, bagging techniques, SMOTE oversampling, and the Voting Classifier, using algorithms such as KNN and SVM, can achieve an ACC of 0.82. Overall, this study proposes a replicable data mining-based workflow that integrates machine learning techniques for predicting honey yield from climatic variables, including the use of an open-access dataset, explainability analysis, and a comparative evaluation of machine learning models, contributing to the development of future analysis and planning tools in the beekeeping sector.","养蜂人需要借助机器学习和物联网等现代技术支持的规划工具，以应对蜂蜜产量下降和波动等农业挑战。为确保可复现性，本文提出了一套研究流程：首先构建一个开放获取数据库，该数据库基于年度生产记录和气候变量（温度和降雨量）开发，并整合了数据构建、可解释性分析和模型评估。随后，应用特征重要性方法和可解释性技术，如特征重要性、随机森林中各特征的深度频率以及Shapley加性解释方法。最后，评估多种机器学习方法用于蜂蜜产量预测：逻辑回归、k近邻、支持向量机、决策树、多层感知机、随机森林、线性判别分析、梯度提升和朴素贝叶斯。这些算法在以下方面进行比较：（1）基线模型，即未进行超参数优化的模型，采用留一交叉验证和分层10折交叉验证；（2）基线加归一化\u002F标准化（最大值除法、最小-最大和z-score）；（3）第2点配置加装袋法；（4）使用SMOTE评估数据增强和类别平衡策略，并结合投票分类器进行模型组合。结果表明，降雨量是蜂蜜产量预测中最重要的变量之一。通过选择特定特征，模型在某些情况下性能得到提升，或性能未显著下降。最小-最大和z-score方法使部分算法的预测效果得到改善；例如，支持向量机达到了0.80的准确率，而参考研究中基于随机森林的准确率为0.67，提升了13个百分点。最后，使用KNN和SVM等算法结合装袋技术、SMOTE过采样和投票分类器，可以达到0.82的准确率。总体而言，本研究提出了一套可复现的、基于数据挖掘的工作流程，整合了机器学习技术以从气候变量预测蜂蜜产量，包括使用开放获取数据集、可解释性分析以及机器学习模型的比较评估，为养蜂领域未来分析和规划工具的开发做出了贡献。","Smart Agricultural Technology","2026-09-20T00:00:00Z",71,{"impact":17,"substance":18,"depth":19,"authority":81,"freshness":20,"relevant":21,"comment":276},"该论文提出可复现的机器学习工作流，结合开放数据集与可解释性方法预测蜂蜜产量，方法新颖、结论可靠，对养蜂业数字化规划有参考价值。",[278],{"name":272,"url":269},[177,26,280,27,281],"产量预测","蜂产业",[283,284],"蜂蜜产量 机器学习 预测","Smart Agricultural Technology 蜂蜜","蜂蜜产量机器学习预测-3258","10.1016\u002Fj.atech.2026.102575",{"doi":286,"openalex_id":288,"authors":289,"venue":272,"cited_by_count":35,"oa_url":269,"card":318,"direction":62,"ingested_from":64},"W7213773430",[290,293,296,299,301,304,307,310,312,315],{"name":291,"orcid":292},"Roberto Ahumada‐García","https:\u002F\u002Forcid.org\u002F0000-0003-1107-4606",{"name":294,"orcid":295},"David Zabala‐Blanco","https:\u002F\u002Forcid.org\u002F0000-0002-5692-5673",{"name":297,"orcid":298},"Víctor Hugo Monzón","https:\u002F\u002Forcid.org\u002F0000-0001-9729-7768",{"name":300,"orcid":9},"Iván Sánchez",{"name":302,"orcid":303},"Nádia Félix Felipe da Silva","https:\u002F\u002Forcid.org\u002F0000-0002-3875-2211",{"name":305,"orcid":306},"Thierson Couto Rosa","https:\u002F\u002Forcid.org\u002F0000-0001-7117-3994",{"name":308,"orcid":309},"Alef Iury Siqueira Ferreira","https:\u002F\u002Forcid.org\u002F0000-0002-9119-6357",{"name":311,"orcid":9},"Xaviera López-Cortés",{"name":313,"orcid":314},"Marco Javier Flores-Calero","https:\u002F\u002Forcid.org\u002F0000-0001-7507-3325",{"name":316,"orcid":317},"Philip Vásquez-Iglesias","https:\u002F\u002Forcid.org\u002F0009-0008-2109-8787",{"tldr":319,"method":320,"finding":321,"direction":62,"opportunity":322},"构建可复现工作流，用气候变量与机器学习分类蜂蜜产量。","开放数据库、特征重要性\u002FSHAP、9种ML模型、SMOTE与投票集成。","降雨是最重要变量；SVM准确率0.80，集成后达0.82。","可扩展至多源物联网数据与实时预测，开发养蜂决策支持工具。","2026-09-23T23:30:03.793669Z"]