基于机器学习理论的土壤侵蚀模型构建

土壤侵蚀一直是环境问题中的重点和难点。由于影响土壤侵蚀的因素众多,传统的预测模型存在数据获取困难、适用范围小、研究周期长等不足,使得对土壤侵蚀的预测无法做到快速、便捷。支持向量机(Support Vector Machine,SVM)是机器学习中的一个重要模型,具有非线性映射、自我学习能力、全局最小值、对输入数据变化不敏感等优点,在建立土壤侵蚀量相关性预测模型方面较传统预测模型具有更强的优势。本研究应用浙江省诸暨市浦阳江水文站的降雨数据,利用ArcGIS地理信息系统确定水文站上游流域为研究区域。以降雨量、研究区域地理数据维度(包括坡度数据、坡长数据、土壤信息、土地利用类型)作为影响因子,输入支持向量机模型,进行流域内土壤侵蚀量预测。将水文站土壤侵蚀量实测数据作为对照值,用模型输出值检验,从而在取值范围内选择出模型最优的参数组。用影响因子数据和土壤侵蚀量数据对使用最优参数的模型进行检验,模型的预测准确率最高达到75%。其中,降雨量对土壤侵蚀量的影响最大,降雨量单因子预测准确率在70%以上,其余因子预测准确率在3.5%左右。最终得到一个土壤侵蚀量相关性预测模型,通过水文站降雨数据以及地理信息,即可预测当地土壤侵蚀量,准确率达到75%。     

基于支持向量机回归的次降雨小流域侵蚀产沙预报研究——以晋西王家沟为例

基于次降雨小流域侵蚀产沙过程的复杂性、非线性，利用支持向量机回归和主成分分析方法，确定了影响次降雨小流域侵蚀产沙量的关键因子，包括浑水径流深、洪峰最大流量、降雨量和30min最大降雨强度。建立了向量机回归支持下的次降雨小流域侵蚀产沙预测模型。利用60次侵蚀产沙实测资料，对模型预报精度进行了分析，结果表明，基于支持向量回归的次降雨流域侵蚀产沙预报模型具有较好的预测精度，预测精度平均为在86％。该研究为揭示次降雨小流域土壤侵蚀规律提供了新的途径和方法。     

基于遗传算法最小二乘支持向量机的耕地变化预测

针对耕地变化内部规律与模拟方法进行研究,提出最小二乘支持向量机耕地变化预测方法,有效构建耕地变化与耕地变化影响因子之间复杂的非线性关系模型。利用遗传算法全局寻优功能优化最小二乘支持向量机内部参数,提高最小二乘支持向量机耕地变化预测模型精度。利用该模型对江苏无锡市1987－2000年期间耕地变化进行预测,并与多元回归、GM(1,1)、BP网络、支持向量机（SVM）耕地预测模型和实际调查耕地变化数据进行比较分析。预测精度评价结果证实,该方法耕地预测精度远高于多元回归、GM(1,1),BP网络模型,略高于SVM模型,但算法复杂度和计算效率远优于SVM预测模型,是一种有效的耕地变化预测方法。     

基于环境减灾卫星高光谱数据的盐碱地等级划分

为了进行盐碱地的有效防治,以松辽盆地为例,基于环境减灾卫星(HJ-1A)高光谱数据,对比曲线回归、最小二乘支持向量机回归二种非线性回归模型在含盐率反演中的预测效果,探索该区土壤盐碱化指标定量反演的最佳模型,最终采用最小二乘支持向量机(LS-SVM)回归预测的方法,在盐碱化较严重的大庆地区进行了多种盐碱地指标反演,并采用...     

应用红外光声光谱技术及支持向量机模型测定土壤有机质含量

快速测定土壤有机质含量对作物生产和土壤肥力评价具有重要意义,红外光声光谱技术的应用为土壤有机质快速测定提供了可能。本研究以江苏省南京市溧水区水稻土土样为材料,探究了红外光声光谱技术在有机质测定中的应用。采用主成分分析、偏最小二乘和独立成分分析,分别提取了土壤光谱的主成分、偏最小二乘潜变量和独立成分,并以提取的信息输入支持向量机,从而构建了三种支持向量机校正模型。同时,偏最小二乘也被用于建立校正模型,作为支持向量机模型的对照。预测结果表明,基于独立成分的支持向量机模型效果最好,预测相关系数R2、均方根误差RMSEP和实际测量值的标准差与光谱模型预测值标准差的比值即RPD值分别为0.808、0.575和2.28。F检验表明,该模型显著优于基于主成分的支持向量机模型,但与基于偏最小二乘潜变量的支持向量机模型,以及经典偏最小二乘模型没有显著差异。t检验表明,各校正模型对有机质的预测结果与化学测定结果没有显著差异。因此,红外光声光谱技术为土壤有机质的快速测定提供了新的技术手段。     

改进支持向量机的股票价格预测研究

为了提高股票价格预测精度,提出一种改进支持向量机的股票价格预测模型。该模型利用粒子群算法的全局寻优能力对支持向量机参数进行优化,以提高股票价格的预测精度,采用具体股票价格数据对模型性能进行测试。结果表明,改进支持向量机能够对股票价变化趋势进行预测,是一种有效、高精度的股票价格预测模型。     

基于高光谱的柑橘叶片钾含量快速诊断模型

及时准确地监测柑橘树体钾营养状况,有助于科学施肥,提高柑橘果实的品质和产量。试验对采集的W·默科特叶片钾含量及叶片光谱信息进行相关性分析,利用主成分分析和连续投影算法进行降维处理,同时结合偏最小二乘回归和最小二乘支持向量机回归分别建立了叶片钾含量预测模型。结果显示,柑橘叶片钾含量特征波长主要集中在450～600 nm、700 nm左右、980～1030 nm区域;最小二乘支持向量机回归模型的预测效果明显优于偏最小二乘回归模型,预测集相关系数达0.91。结果表明,利用高光谱成像技术结合最小二乘支持向量机回归可实现对柑橘叶片钾含量的快速诊断。     

基于细菌觅食优化算法的支持向量机在土壤墒情预测中的应用

[目的]对基于细菌觅食优化算法的支持向量机在土壤墒情预测中的应用进行探讨,为现代农业研究中土壤墒情预测及农业生产提供支持。[方法]基于支持向量回归机方法建立土壤墒情预测模型,利用细菌觅食优化算法优化支持向量机预测模型的相关参数。根据从种植区采集的田间数据对模型进行建模和测试。[结果]与仅利用支持向量回归机和利用粒子群优化的支持向量回归机分别建立的模型进行对比,发现本研究所提算法建立的预测模型的预测效果更佳。[结论]该模型预测效果较好,所建模型已应用于实际项目,预测精度基本满足要求,且运行稳定。进而证明了该研究所提算法的有效性和可行性。     

支持向量机、信息向量机和相关向量机的比较研究

介绍了支持向量机、信息向量机和相关向量机的理论与算法。利用最优化对偶理论,阐述了支持向量机的三种主要算法：硬间隔支持向量机、软间隔线性支持向量机和二次软间隔支持向量机的理论推导过程。对基于高斯过程模型,详细说明了信息向量机和相关向量机算法的实现过程。     

基于最小二乘向量机土壤水分动态模拟与分析

土壤水分动态的模拟对水分循环与农业生产中水分的合理利用与管理具有重要的意义.应用最小二乘支持向量机对加入气象因子随机变量的红壤中土壤水分动态变化进行了训练、检验及模拟.结果表明,最小二乘支持向量机相比与神经网络方法不论是模拟性能指标还是建模的数学意义都有更好的可靠性和优越性;本研究应用最小二乘支持向量机对土壤水分动态日变化进行了模拟,并采用bior 3.3小波函数5层分解提取日变化趋势图进而把该研究区土壤水分日变化划分为4个阶段,其结果可为研究区水分合理利用和土壤墒情的预测预报提供科学依据.     

多分类支持向量机在泥石流危险性区划中的应用

以凉山州安宁河流域129个乡镇的泥石流危险性区划资料为依据,随机选取总样本数的2/3和1/2作为训练样本,建立不同数量训练样本下安宁河流域泥石流危险性区划的多分类SVM模型,进行以乡镇为单元的区域泥石流危险性评价研究。评价结果表明,SVM模型的预测精度随着训练样本数量的增加而提高;2个SVM模型对测试样本的预测准确率均高于相应的BP神经网络模型,对训练样本的回判准确率高于或接近于BP神经网络模型。因此,支持向量机方法是一种比神经网络方法具有更优精度和更强泛化性能的新机器学习方法,在泥石流危险性评价实践中具有十分广阔的应用前景和推广应用价值。     

Assessment of gully erosion susceptibility using different DEM-derived topographic factors in the black soil region of Northeast China

As a primary sediment source, gully erosion leads to severe land degradation and poses a threat to food and ecological security. Therefore, identification of susceptible areas is critical to the prevention and control of gully erosion. This study aimed to identify areas prone to gully erosion using four machine learning methods with derived topographic attributes. Eight topographic attributes (elevation, slope aspect, slope degree, catchment area, plan curvature, profile curvature, stream power index, and topographic wetness index) were derived as feature variables controlling gully occurrence from digital elevation models with four different pixel sizes (5.0 m, 12.5 m, 20.0 m, and 30.0 m). A gully inventory map of a small agricultural catchment in Heilongjiang, China, was prepared through a combination of field surveys and satellite imagery. Each topographic attribute dataset was randomly divided into two portions of 70% and 30% for calibrating and validating four machine learning methods, namely random forest (RF), support vector machines (SVM), artificial neural network (ANN), and generalized linear models (GLM). Accuracy (ACC), area under the receiver operating characteristic curve (AUC), root mean square error (RMSE), and mean absolute error (MAE) were calculated to assess the performance of the four machine learning methods in predicting spatial distribution of gully erosion susceptibility (GES). The results suggested that the selected topographic attributes were capable of predicting GES in the study catchment area. A pixel size of 20.0 m was optimal for all four machine learning methods. The RF method described the spatial relationship between the feature variables and gully occurrence with the greatest accuracy, as it returned the highest values of ACC (0.917) and AUC (0.905) at a 20.0 m resolution. The RF was also the least sensitive to resolutions, followed by SVM (ACC = 0.781–0.891, AUC = 0.724–0.861) and ANN (ACC = 0.744–0.808, AUC = 0.649–0.847). GLM performed poorly in this study (ACC = 0.693–0.757, AUC = 0.608–0.703). Based on the spatial distribution of GES determined using the optimal method (RF + pixel size of 20.0 m), 16% of the study area has very high level susceptibility classes, whereas areas with high, moderate, and low levels of susceptibility make up approximately 24%, 30%, and 31% of the study area, respectively. Our results demonstrate that GES assessment with machine learning methods can successfully identify areas prone to gully erosion, providing reference information for future soil conservation plans and land management. In addition, pixel size (resolution) is the key consideration when preparing suitable datasets of feature variables for GES assessment.     

结合光谱降维的IPSO-SVR水体总磷浓度预测模型

[目的] 选择最优模型对水体中总磷浓度进行预测,为准确、实时、高效检测水资源状况提供支持。 [方法] 以2021年在长江中下游武汉—安徽地区采集的水质样本作为研究对象,首先,对采集到的长江光谱数据进行最大最小归一化和均值中心化两种预处理操作以便统一数据的范围和均值点,并使用核主成分分析(KPCA)技术对预处理后的光谱数据进行降维操作。选取方差解释率为99.6%下的6个特征向量进行后续预测模型的训练,接着在原有粒子群算法的基础上引入自适应惯性权重更新公式和遗传—模拟退火变异思想,提高算法的寻优能力。使用改进的粒子群优化算法对支持向量回归模型中的超参数组合进行寻优,对支持向量回归模型使用输出的结果进行预测模型的训练,最后使用测试集数据进行总磷浓度的预测。 [结果] 提出了一种结合光谱降维的改进粒子群优化算法(IPSO)结合支持向量回归(SVR)的水体总磷含量预测模型。通过和当前预测性能较好的几种机器学习模型进行精度的比较发现,该试验模型对长江水体总磷浓度进行预测时决定系数(R²)为0.973 920,均方根差(RMSE)为0.003 012,平均绝对误差(MAE)为0.002 105。 [结论] 使用光谱数据结合降维技术、粒子群优化算法和机器学习模型的算法融合模型检测水体总磷浓度可行性强,精确度高,且拟合效果良好。     

Supervised Machine Learning Algorithms Can Predict Penetration Resistance in Mineral-fertilized Soils

ABSTRACT

It is desirable for farmers and agricultural engineers to estimate the changes in soil penetration resistance (SPR) due to the usage of mineral fertilizers. However, reliable instruments for such measurements are not always available. In this study, the effects of conventional N, P, and K-based mineral fertilizers are firstly investigated on SPR. Then, machine learning methods are utilized to predict SPR. The cone index as an indicator of SPR was measured in different soil depths between 0 and 90 cm, different soil moisture contents from 10% to 30%, and different soil organic matter contents from 1% to 8%. Experimental data showed that fertilizing treatments were more effective on the SPR of soil samples (p < .05) in higher penetration depths. Double fertilizing of urea resulted in the highest SPR in the studied range of soil depth. Fertilizing treatments, soil moisture contents, soil organic matter contents, and penetration depths as input features of the samples were then trained to the machine using supervised learning methods and the machine’s performance in prediction of SPR was investigated. Results showed that support vector machine resulted in an acceptable performance, normalized MSE and R² of which were equal to 0.009 and 0.98, respectively using Gaussian kernel with kernel parameter of 30. Findings of this study reveal that machine learning can reliably predict SPR when instruments such as cone penetrometer are not available.     

A comparative study: Prediction of constructed treatment wetland performance with k-nearest neighbors and neural networks

K-nearest neighbors (KNN), support vector machine (SVM) and self-organizing map (SOM) were applied to predict five-day @ 20?C N- Allylthiourea biochemical oxygen demand (BOD) and suspended solids (SS), and to assess novel alternative methods of analyzing water quality performance indicators for constructed treatment wetlands. Concerning the accuracy of prediction, SOM showed a better performance compared to both KNN and SVM. Moreover, SOM had the potential to visualize the relationship between complex biochemical variables. However, optimizing the SOM requires more time in comparison to KNN and SVM because of its trial and error process in searching for the optimal map. The results suggest that BOD and SS can be efficiently estimated by applying machine learning tools with input variables such as redox potential and conductivity, which can be monitored in real time. Their performances are encouraging and support the potential for future use of these models as management tools for the day-to-day process control.     

An Insight into Machine Learning Algorithms to Map the Occurrence of the Soil Mattic Horizon in the Northeastern Qinghai-Tibetan Plateau

Soil diagnostic horizons, which each have a set of quantified properties, play a key role in soil classification. However, they are difficult to predict, and few attempts have been made to map their spatial occurrence. We evaluated and compared four machine learning algorithms, namely, the classification and regression tree(CART), random forest(RF), boosted regression trees(BRT), and support vector machine(SVM), to map the occurrence of the soil mattic horizon in the northeastern Qinghai-Tibetan Plateau using readily available ancillary data. The mechanisms of resampling and ensemble techniques significantly improved prediction accuracies(measured based on area under the receiver operator characteristic curve score(AUC)) and produced more stable results for the BRT(AUC of 0.921 ± 0.012, mean ± standard deviation) and RF(0.908 ± 0.013) algorithms compared to the CART algorithm(0.784 ± 0.012), which is the most commonly used machine learning method. Although the SVM algorithm yielded a comparable AUC value(0.906 ± 0.006) to the RF and BRT algorithms, it is sensitive to parameter settings, which are extremely time-consuming.Therefore, we consider it inadequate for occurrence-distribution modeling. Considering the obvious advantages of high prediction accuracy, robustness to parameter settings, the ability to estimate uncertainty in prediction, and easy interpretation of predictor variables, BRT seems to be the most desirable method. These results provide an insight into the use of machine learning algorithms to map the mattic horizon and potentially other soil diagnostic horizons.     

基于多目标优化方法的滑坡易发性评价

[目的] 在滑坡易发性评价中,滑坡预测模型的选取和优化对运算过程的高效性和预测结果的准确性至关重要。针对现有单目标遗传优化算法(genetic algorithm,GA)易陷入早熟、局部搜索能力差、全局优化速度慢等问题,拟提出一种新的优化算法框架,将多目标遗传算法中的经典算法—带精英选择策略的非支配排序算法(the nondominated sorting genetic algorithm with an elite strategy,NSGA-Ⅱ)与常用机器学习模型[随机森林(random forest,RF)、支持向量机(support vector machine,SVM)]相结合,进行滑坡易发性预测。与单目标优化不同的是,NSGA-Ⅱ算法可同时进行特征选择和超参数优化,并使预测模型同时实现最优准确度、召回率、精密度和AUC(area under curve,AUC)。[方法] 以三峡库区重庆段为研究区,从模型精度评价、滑坡灾害易发性分区图、分区统计3个方面对4种优化模型(RF-GA、SVM-GA、RF-NSGA-II、SVM-NSGA-II)进行对比分析。[结果] NSGA-II较GA优化效果更明显,在模型评价和滑坡易发性分区方面,RF-NSGA-II模型具有更高的预测性能,4项评价值分别为80.91%,81.89%,80.07%,88.60%,证明NSGA-II优化算法的有效性;极低至极高危险区面积占比依次为23.06%,22.46%,22.96%,19.99%,11.53%,验证了RF-NSGA-II模型的可靠性。由RF-NSGA-II模型预测得到的易发性图表明,高和极高易发性区集中在研究区北部,且由东向西呈带状分布。[结论] 研究采取的基于多目标选择的RF-NSGA-II模型,为滑坡易发性评价中机器学习模型调优提供新思路。     

基于多源信息融合技术的马铃薯痂疮病无损检测方法

为了提高马铃薯痂疮病无损检测识别精度,基于机器视觉和近红外光谱的多源信息融合技术,该文提出DS（dempster shafer）证据理论结合支持向量机的马铃薯痂疮病无损检测方法。试验以360个马铃薯为研究对象,在图像特征分割时,确定了差影法结合马尔可夫随机场模型法为最佳分割方法;在光谱特征提取时,确定主成分分析方法为最佳降维方法。采用支持向量机识别方法分别建立机器视觉和近红外光谱的马铃薯痂疮病识别模型,模型对测试集马铃薯识别率分别为89.17%、91.67%。采用DS证据理论与支持向量机相结合的方法对获取的图像特征和光谱特征进行融合,建立了基于机器视觉和近红外光谱技术的多源信息融合马铃薯痂疮病检测模型,该模型对测试集马铃薯识别率为95.83%。试验结果表明,该技术对马铃薯痂疮病进行检测是可行的,融合模型比单一的机器视觉模型或近红外光谱模型识别率高。     

基于支持向量回归（SVR）和多时相遥感数据的冬小麦估产

发展和构建高精度的作物遥感估产模型,对于国家制订粮食进出口政策和保障粮食安全具有重要意义.尝试利用支持向量回归方法(SVR)构建遥感估产模型,首先利用北京郊区2004年和2007年冬小麦主要生育期多时相Landsat TM影像生成的归一化植被指数,通过SVR构建遥感估产模型进行产量估算.然后针对模型的稳健型和预报能力进行交叉验证,并与常规的多元回归方法进行对比.结果表明,利用SVR方法构建的遥感估算模型有效地提高了估算精度,与多元回归方法相比,2004年和2007年决定系数分别提高0.2162、0.2158,均方根误差分别降低0.1682、0.2912.因此基于SVR和多时相遥感数据构建估产模型用于冬小麦估产是可行、有效的,为应用多时相遥感数据进行冬小麦估产提供了一种方法.     

基于最小二乘支持向量机的中国粮食产量预测模型研究

粮食产量预测是制定农业政策的重要依据。针对农业生产系统的特征,在统计学习理论和结构风险最小化原理的基础上,建立了基于最小二乘支持向量机的时间预测模型。预测结果表明该模型具有较高的预测精度,为粮食产量预测提供了一条新的途径。