Modeling wheat yield with data-intelligent algorithms: artificial neural network versus genetic programming and minimax probability machine regression

doi:10.1016/B978-0-12-816514-0.00002-3

Handbook of Probabilistic Models

2020, Pages 37-87

https://doi.org/10.1016/B978-0-12-816514-0.00002-3 Get rights and content

Abstract

In precision agriculture, data-intelligent algorithms applied for predicting wheat yield can generate crucial information about enhancing crop production and strategic decision-making. In this chapter, artificial neural network (ANN) model is trained with three neighboring station-based wheat yields to predict the yield for two nearby objective stations that share a common geographic boundary in the agricultural belt of Pakistan. A total of 2700 ANN models (with a combination of hidden neurons, training algorithm, and hidden transfer/output functions) are developed by trial-and-error method, attaining the lowest mean square error, in which the 90 best-ranked models for 3-layered neuronal network are utilized for wheat prediction. Models such as learning algorithms comprised of pure linear, tangent, and logarithmic sigmoid equations in hidden transfer/output functions, executed by Levenberg–Marquardt, scaled conjugate gradient, conjugate gradient with Powell-Beale restarts, Broyden–Fletcher–Goldfarb–Shanno quasi-Newton, Fletcher-Reeves update, one-step secant, conjugate gradient with Polak-Ribiére updates, gradient descent with adaptive learning, gradient descent with momentum, and gradient descent with momentum adaptive learning method are trained. For the predicted wheat yield at objective station 1 (i.e., Toba Taik Singh), the optimal architecture was 3-14-1 (input-hidden-output neurons) trained with the Levenberg–Marquardt algorithm and logarithmic sigmoid as activation and tangent sigmoid as output function, while at objective station 2 (i.e., Bakkar), the Levenberg–Marquardt algorithm provided the best architecture (3-20-1) with pure liner as activation and tangent sigmoid as output function. The results are benchmarked with those from minimax probability machine regression (MPMR) and genetic programming (GP) in accordance with statistical analysis of predicted yield based on correlations (r), Willmott's index (WI), Nash-Sutcliffe coefficient (EV), root mean-squared error (RMSE), and mean absolute error (MAE). For objective station 1, the ANN model attained the r value of approximately 0.983, with WI ≈ 0.984 and EV ≈ 0.962, while the MPMR model attained r ≈ 0.957, WI ≈ 0.544, and EV ≈ 0.527, with the results attained by GP model, r ≈ 0.982, WI ≈ 0.980, and EV ≈ 0.955. For optimal ANN model, a relatively low value of RMSE ≈ 192.02 kg/ha and MAE ≈ 162.75 kg/ha was registered compared with the MPMR (RMSE ≈ 614.46 kg/ha; MAE ≈ 431.29 kg/ha) and GP model (RMSE ≈ 209.25 kg/ha; MAE ≈ 182.84 kg/ha). For both objective stations, ANN was found to be superior, as confirmed by a larger Legates-McCabe's (LM) index used in conjunction with relative RMSE and MAE. Accordingly, it is averred that ANN is considered as a useful data-intelligent contrivance for predicting wheat yield by using nearest neighbor yield.

References (0)

Cited by (16)

Predicting the small strain shear modulus of sands and sand-fines binary mixtures using machine learning algorithms
2024, Transportation Geotechnics
This study aims to develop several novel machine learning (ML) evolutionary algorithms for the prediction of small strain shear modulus ( $G_{\max}$ ) of clean sands and sand-fines binary mixtures. To this end, five key features of isotropic confining pressure ( $p$ ), void ratio ( $e$ ), uniformity coefficient ( $C_{u}$ ), particle shape descriptor ( $ρ$ ), and non-plastic fines content ( $FC$ ) are adopted as the inputs to artificial neural network (ANN) models as well as genetic programming (GP) algorithm so as to render the maximum shear modulus of granular soils as the output. Accordingly, a comprehensive dataset containing 1055 $G_{\max}$ data points is exploited to develop ML simulations. The validity of ML-based models in estimating the $G_{\max}$ of clean sands and sand-silt mixtures is rigorously examined through various statistical indices and measurement criteria. The results show that a novel ML model utilizing ANN-Levenberg Marquardt (LM) in conjunction with an evolutionary optimization method named Success History-based Adaptive Differential Evolution with Linear population size reduction (LSHADE) is capable of predicting $G_{\max}$ data with a very high precision rendering R² values of 0.9833, 0.9841, 0.9802, and 0.9835 for the whole, training, validation, and test datasets, respectively. Meanwhile, using the well-established GP algorithm, a new practical model is proposed to predict the $G_{\max}$ of clean sands and sand-fines mixtures containing non-cohesive silt inclusion with R² values of 0.9323, 0.9351, and 0.9312 for the whole, training, and test datasets, respectively. Finally, the proposed models of ANN-LSHADE-LM and GP are shown to be appreciably superior, in terms of accuracy, to all commonly used empirical correlations in the literature for $G_{\max}$ estimation.
Optimal configuration of architectural building design parameters for higher educational buildings
2023, Energy Reports
In architectural building design, finding an equilibrium between energy consumption and thermal comfort poses a challenge due to their inherent conflict. To effectively incorporate both aspects, their integration into the decision-making process is crucial. This study demonstrates the optimization of architectural building design parameters for a single-story educational building within a warm temperate climate in New South Wales, Australia. A meticulous validation process was undertaken using the Monte Carlo approach and scrutinizing 2,000 distinct scenarios for each parameter. This validation compared simulation outcomes with field-measured data, confirming the model’s prediction accuracy. The optimization endeavour employed the non-dominated sorting genetic algorithm to facilitate multiobjective optimization , working with the Pareto front solution. When compared with the baseline building simulation model, the optimized architectural design parameter configuration yields significant reductions of up to 24% in total EC and up to 16.5% in thermal discomfort hours.
Soft computing approach on estimating the lateral confinement coefficient of CFRP veiled circular columns
2023, Alexandria Engineering Journal
For any structures, Reinforced concrete (RC) columns are the utmost carping members. RC columns jacketed with CFRP membrane enhance resistance, deformation aptitude, and the auspicious confining impacts. Sundry scrutiny has been implemented about circular columns wrapping or veiled with CFRP layers. This scrutiny will manifest about Genetic Programming (GP) and Artificial Neural Networks (ANN) as soft computing models, mobbing exploited to ascertain the lateral confinement coefficient (K_s) valuate of RC circular columns jacketed with CFRP membrane. For the purpose to ascertain K_s value, manipulate the corresponding ambit such as the diameter of the column (b), column length (L), thickness layer of CFRP (t_w), the elastic modulus of CFRP membrane (E_f) along with the compressive strength of unconfined concrete (f_cc). Total collection of 142 datasets for the RC circular column was employed to decipher the prediction of K_s. Various statistical metrics (R² = 0.933; RMSE = 0.054; NMBE = -0.001) compared with the ANN model and exhibits the superiority of GP in predicting the K_s. Taylor diagram and Rank analysis were also carried out in this article for justification. The lucidity of the GP machine enlightenment model overtures a greater sensible and exact prediction of K_s value for RC circular columns in accordance with versatile loading conditions.
A comparative of modeling techniques and life cycle assessment for prediction of output energy, economic profit, and global warming potential for wheat farms
2022, Energy Reports
Uncertainty about the energy use efficiency, lack of knowledge about economic outcomes, and its environmental consequences have always take risks in changing cultivation patterns and moving towards the optimal path. Accordingly, this study provided mathematical, artificial neural networks (ANNs), adaptive neuro-fuzzy inference system (ANFIS) methods to predict output energy, economic profit, and global warming potential (GWP) of wheat production. For this purpose, 75 wheat farms located in the central area of Hamadan province, Iran, were selected randomly, and data were gathered through oral interviews. After collecting input and output energies data, the averages of inputs and outputs energies were obtained about 43055 MJ ha⁻¹ and 117407 MJ ha⁻¹, respectively. Economic analysis has performed in the next step. Its results revealed that the benefit-to-cost ratio and net return were computed about 2.33 and 488.29 $ per ha for wheat production. Then, life cycle assessment (LCA) was utilized to specify the environmental effects of wheat cultivation, and its results demonstrated that GWP is the most important environmental impact which caused 624.29 kg CO₂ $_{eq.}$ during 1 ton of wheat production. Modeling results illustrated R² was varied between 0.264 and 0.978 in the linear regression, 0.313 and 954 in the best structure of ANN with two hidden layers, and 0.520 and 0.962 in the ANFIS with three-level structure. Modeling comparison indicated that generally, ANFIS model with considering all uncertainty items can be offered better prediction models among all and after that ANN with considering non-linear parameters is in the next rank.
Using soft computing and leaf dimensions to determine sex in immature Pistacia vera genotypes
2021, Measurement: Journal of the International Measurement Confederation
Citation Excerpt :
Soft computing such as use of neural networks has hastened our ability to predict physiological processes within living organs [13,14]. Methods like K-nearest neighbor (KNN), support vector machine (SVM), self-organizing map (SOM), artificial neural network fuzzy inference system (ANFIS), and radial basis function (RBF) are being used for unique applications in animal behavior [15], medicine and cancer diagnosis [16,17], plant species hierarchical classification [18], disease diagnosis [19], image classification [20,21], regression studies [22], prediction of plant yield and performance [21,23,24], tree phenology [25], and predictions of superior genotype [26]. Machine learning based on visual assessment of biometric data has been used for sex determination in humans, fish and other animals [27–29], but has yet to be used in plants.
Pistachio (Pistacia vera) is a dioecious tree species in the cashew family originating from Central Asia and the Middle East. The sex of pistachio genotypes cannot be established until inflorescence growth during the reproductive stage. The ability to determine sex before flowering would be extremely beneficial for pistachio growers and breeders. In this study, different methods of soft computing, using classifiers (i.e. KNN, SVM, SOM, ANFIS and RBF) were employed to determine pistachio sex based on geometric features of mature leaves. Data were collected from more than 900 leaves representing 45 pistachio genotypes before sexual maturation. Separate training data set sizes (20, 40, 60 and 80%) were used to evaluate the generalizability of each classifier. Results indicated the KNN classifier was the most accurate (>95%) for sex determination during training phases however, accuracy during testing phases varied from 52 to 60% for the four training data set sizes. Training phase accuracy with the 80% data set ranged from 83 to 64% with the respective arrangement of RBF, ANFIS, SVM, and the SOM classifiers. In the test phase, classifier accuracy varied from 88.33% (RBF) to 60% (KNN). Our results indicated RBF could reliably be used to differentiate between male and female pistachio genotypes. Thus, soft computing models are useful tools for predicting sex in pistachio based on leaf dimensions.
Reducing deep learning network structure through variable reduction methods in crop modeling
2021, Artificial Intelligence in Agriculture
Citation Excerpt :
Increases in crop productivity have been largely attributed (50–60%) to breeding and the development of hybrid cultivars, followed by improved management practices (Connor et al., 2011; Sacks and Kucharik, 2011). Technological advancements have helped fine-tune management with increased adaptation of yield simulation (Ali and Deo, 2020; Ali et al., 2018a), field monitoring (Rao and Sridhar, 2018), and other data-driven practices (Pathak et al., 2018). Most recently, precision in resource management has been continuously fine-tuned through crop models and use of satellite navigation systems (Abbasi et al., 2014; Basso et al., 2001; Lobell and Burke, 2010).
Crop models are widely used to predict plant growth, water input requirements, and yield. However, existing models are very complex and require hundreds of variables to perform accurately. Due to these shortcomings, large-scale applications of crop models are limited. In order to address these limitations, reliable crop models were developed using a deep neural network (DNN) – a new approach for predicting crop yields. In addition, the number of required input variables was reduced using three common variable selection techniques: namely Bayesian variable selection, Spearman's rank correlation, and Principal Component Analysis Feature Extraction. The reduced-variable DNN models were capable of estimating future crop yields for 10,000,000 different weather and irrigation scenarios while maintaining comparable accuracy levels to the original model that used all input variables. To establish clear superiority of the methodology, the results were also compared with a very recent feature selection algorithm called min-redundancy max-relevance (mRMR). The results of this study showed that the Bayesian variable selection was the best method for achieving the aforementioned goals. Specifically, the final Bayesian-based DNN model with a structure of 10 neurons in 5 layers performed very similarly (78.6% accuracy) to the original DNN crop model with 400 neurons in 10 layers, even though the size of the neural network was reduced by 80-fold. This effort can help promote sustainable agricultural intensifications through the large-scale application of crop models.

View all citing articles on Scopus

View full text

Chapter 2 - Modeling wheat yield with data-intelligent algorithms: artificial neural network versus genetic programming and minimax probability machine regression

Abstract