Optimizing Lettuce Crop Growth Modeling with XGBoost-SVM and Gaussian Process Regression Fusion

Agriculture plays a pivotal role in ensuring food security and sustainable development. Accurate prediction of crop growth and yield is essential for optimizing agricultural practices and resource allocation. This study presents an integrated approach employing various machine learning techniques to predict crop growth, exemplified by a dataset of 70 plant samples. The dataset includes environmental factors such as temperature, humidity, soil composition, and pH levels, as well as temporal information like growth days. The dataset goes through a thorough preprocessing step that includes resolving missing values, eliminating duplicates, and, encoding categorical variables. To make sure that every variable contributes equally to the models, feature scaling approaches like Z-score standardization and Min–Max scaling are used. The probabilistic outputs of the SVM and XGBoost models are combined in an ensemble technique to improve prediction accuracy. This fusion strategy utilizes the strengths of each model and reduces individual model biases. In addition to classification models, Gaussian Process Regression (GPR) is utilized to provide probabilistic crop growth predictions. GPR offers a continuous output that can be used for yield estimation and uncertainty quantification. Model performance is evaluated using evaluation measures including R-squared (R2), Mean Absolute Error (MAE), and Root Mean Squared Error (RMSE). These measures offer information on the precision, accuracy, and capacity of the models to explain crop growth variance. This work suggests using Gaussian Process Regression, a machine learning technique, to improve the outcomes of crop growth monitoring for lettuce in aeroponic vertical farming systems. Compared to previous models, this one generates higher prediction results of 0.99% and has lower error rates of 0.32 (MSE), 0.56 (RMSE), and 0.20 (MAE). It is also very resilient to outliers.