Multivariate Linear Regression Model for Predicting Soil Texture and Plant Degradation in Parks

Abstract

The soil, which makes up the top layer of the Earth’s surface, is crucial for the energy and nutrient flows required for the growth of plants and is, therefore, crucial for the environmental study of parks. Since plants and soil are interdependent, soil degradation affects plant quality. The soil’s capacity to store water and support germination directly correlates with plant growth, and soil degradation ultimately results in plant death. One of the most crucial elements in preventing soil degradation is maintaining ecological factors and identifying their impact on plant life through relationships. Therefore, it is essential to simultaneously evaluate and predict the overall condition of plants and soil in designated regions like parks and reserved areas. Hence a model is developed to predict park management practices like Soil Texture (ST) and its impacts on Plants Degradation (PD) using Multivariate Linear regression (MvLR-STPD) to maintain the park’s ecology. Initially, creating a predictive model that can precisely estimate soil texture and plant degradation in parks based on a collection of predictor variables is the main goal of the MvLR. Secondly, based on the data and the precise correlations between the predictor factors and the targeted variables, gradient descent optimization is used to adapt and modify its parameters. The model can extract the correct value, identify the basic variances, and determine the relationship between the impacted variables owing to the optimization method. The model’s performance is validated by prediction metrics like sensitivity analysis and coefficient of determination R2 with loss function, namely, Mean Squared Error (MSE), Root Mean Squared Error (RMSE), Prediction accuracy, correlation coefficient, and Mean Absolute Error (MAE).