GSA-based support vector neural network: a machine learning approach for crop prediction to provision sustainable farming

Abstract

PurposeAutomated crop prediction is needed for the following reasons: First, agricultural yields were decided by a farmer's ability to work in a certain field and with a particular crop previously. They were not always able to predict the crop and its yield solely on that idea alone. Second, seed firms frequently monitor how well new plant varieties would grow in certain settings. Third, predicting agricultural production is critical for solving emerging food security concerns, especially in the face of global climate change. Accurate production forecasts not only assist farmers in making informed economic and management decisions but they also aid in the prevention of famine. This results in farming systems’ efficiency and productivity gains, as well as reduced risk from environmental factors.Design/methodology/approachThis research paper proposes a machine learning technique for effective autonomous crop and yield prediction, which makes use of solution encoding to create solutions randomly, and then for every generated solution, fitness is evaluated to meet highest accuracy. Major focus of the proposed work is to optimize the weight parameter in the input data. The algorithm continues until the optimal agent or optimal weight is selected, which contributes to maximum accuracy in automated crop prediction.FindingsPerformance of the proposed work is compared with different existing algorithms, such as Random Forest, support vector machine (SVM) and artificial neural network (ANN). The proposed method support vector neural network (SVNN) with gravitational search agent (GSA) is analysed based on different performance metrics, such as accuracy, sensitivity, specificity, CPU memory usage and training time, and maximum performance is determined.Research limitations/implicationsRather than real-time data collected by Internet of Things (IoT) devices, this research focuses solely on historical data; the proposed work does not impose IoT-based smart farming, which enhances the overall agriculture system by monitoring the field in real time. The present study only predicts the sort of crop to sow not crop production.Originality/valueThe paper proposes a novel optimization algorithm, which is based on the law of gravity and mass interactions. The search agents in the proposed algorithm are a cluster of weights that interact with one another using Newtonian gravity and motion principles. A comparison was made between the suggested method and various existing strategies. The obtained results confirm the high-performance in solving diverse nonlinear functions.

Learning latent dynamics with a grey neural ODE prediction model and its application

“…Apart from the commercial and energy sectors, other examples include 1) the medical field (Seneviratna and Rathnayaka, 2022), to predict the evolution of an epidemic using only the number of reported cases; 2) Human geography (Xie et al , 2018), for demographic forecasting without resorting to variables such as activity rate, birth rate, etc. ; 3) Agriculture (Ashwitha and Latha, 2022), for forecasting crop yields independently of the knowledge of multiple associated variables, and several other sectors. In general, the NeuralODE-GM(1,1) model solves practical problems where very complex [1] and huge dataset is needed, allowing users to do without costly forecasting software.…”

The routledge handbook of catalysts for a sustainable circular economy

“…The most commonly used AI/ML models in literature include logistic regression (LR), artificial neural networks (ANN) and deep learning (DL), support vector machines (SVM), decision trees (DT), Bayesian networks (BN), K-nearest neighbours (KNN), gradient boosting machines (GBM), other general models (e.g., random forest and cluster analysis), and hybrid models (Ahmed & Kim, 2017;Akanbi et al, 2019;Ashwitha & Latha, 2022;Bala, 2010;Chachdi et al, 2019;Chakraborty et al, 2020;Gaur et al, 2015;Kharfan et al, 2021;Matino et al, 2019;Nassif, 2016;Rahman et al, 2011;Wang & Dowling, 2022;Wang & Liu, 2021). Generally, AI/ ML models are a range of predictive programming algorithms that use the known to deduce the unknown.…”