Department of Electrical and Computer Engineering, Texas A&M University, College Station, TX, 77843, USA.
Department of Biological and Agricultural Engineering, Texas A&M University, College Station, TX, 77843, USA.
Sci Rep. 2024 Oct 17;14(1):24337. doi: 10.1038/s41598-024-74376-7.
Recent advancements in sustainable agriculture have spurred interest in hydroponics as an alternative to conventional farming methods. However, the lack of data-driven approaches in hydroponic growth presents a significant challenge. This study addresses this gap by varying nitrogen, magnesium, and potassium concentrations in hydroponically grown soybeans and conducting essential nutrient profiling across the growth cycle. Statistical techniques like Linear Interpolation are employed to interpolate nutrient data and a feature selection pipeline consisting of chi-squared testing methods, Linear Regression with Recursive Feature Elimination (RFE) and ExtraTreesClassifier have been used to select important nutrients for predicting water uptake using non-parametric regression methods. For different nutrient growth media, i.e. for soybeans grown in Hoagland + Nitrogen and Hoagland + Magnesium media, the Random Forest regressor outperformed other methods in predicting water uptake, achieving testing Mean Squared Error (MSE) scores of 24.55 ( score 0.63) and 8.23 ( score 0.81), respectively. Similarly, for soybeans grown in Hoagland + Potassium media, Support Vector Regression demonstrated superior performance with a testing MSE of 4.37 and score of 0.85. SHapley Additive exPlanations (SHAP) values are examined in each case to elucidate the contributions of individual nutrients to water uptake predictions. This research aims to provide data-driven insights to optimize hydroponic practices for sustainable food production.
最近可持续农业的进展激发了人们对水培作为传统农业方法替代方案的兴趣。然而，水培生长中缺乏数据驱动的方法是一个重大挑战。本研究通过改变水培大豆中的氮、镁和钾浓度，并在整个生长周期内进行基本营养分析，解决了这一差距。线性插值等统计技术用于插值营养数据，特征选择管道包括卡方检验方法、带递归特征消除 (RFE) 的线性回归和 ExtraTreesClassifier，用于使用非参数回归方法选择预测水分吸收的重要养分。对于不同的养分生长介质，即在 Hoagland+氮和 Hoagland+镁介质中生长的大豆，随机森林回归器在预测水分吸收方面优于其他方法，在测试中分别达到了 24.55（ 得分 0.63）和 8.23（ 得分 0.81）的均方误差 (MSE) 分数。同样，对于在 Hoagland+钾介质中生长的大豆，支持向量回归表现出更好的性能，测试 MSE 为 4.37， 得分为 0.85。在每种情况下，都检查了 Shapley 加法解释 (SHAP) 值，以阐明个别养分对水分吸收预测的贡献。本研究旨在为优化水培实践以实现可持续粮食生产提供数据驱动的见解。
