Advanced machine learning model for better prediction accuracy of soil temperature at different depths

Abstract

Soil temperature has a vital importance in biological, physical and chemical processes of terrestrial ecosystem and its modeling at different depths is very important for land-atmosphere interactions. The study compares four machine learning techniques, extreme learning machine (ELM), artificial neural networks (ANN), classification and regression trees (CART) and group method of data handling (GMDH) in estimating monthly soil temperatures at four different depths. Various combinations of climatic variables are utilized as input to the developed models. The models' outcomes are also compared with multi-linear regression based on Nash-Sutcliffe efficiency, root mean square error, and coefficient of determination statistics. ELM is found to be generally performs better than the other four alternatives in estimating soil temperatures. A decrease in performance of the models is observed by an increase in soil depth. It is found that soil temperatures at three depths (5, 10 and 50 cm) could be mapped utilizing only air temperature data as input while solar radiation and wind speed information are also required for estimating soil temperature at the depth of 100 cm.

Machine Learning-Based Microclimate Model for Indoor Air Temperature and Relative Humidity Prediction in a Swine Building

“…Unlike other ML models that are considered as a black-box model while operation, decision tree regression (DTR) models are own opposite characteristics among the other models. Compared to the other supervised algorithms, DTR is popular for the self-explanatory/rule-based by nature; data interpretability for a response subject to the predictor variables could formulate visually [ 11 ]. DTR models were initially developed to solve the classification problem and manipulated to solve the classification and regression problem (CAR).…”

“…The information gain is applied to calculate the amount of an attribute, which contributes to estimating the classes. The entropy and information gain can be expressed by the following Equations (2) and (3) [ 11 , 29 , 30 ], where p Ti is the proportion of data points; C T is the total number of classes; T i is the one sample among all the n subsets in which the total amount of training data T was divided due to an attribute X.…”

“…Though the ML models have been relatively efficient and popular in recent decades, training methods and the amount of feeding data have contributed to their success. More often researchers used 70:30 (training:validation), 80:20, or 90:10 partition to simulate the models [ 11 , 13 , 27 , 42 , 43 ]. The data partition scale for training and testing to be given during the simulation is assumed to be still unexplained and without any principled reason-based calculation.…”

“…The results of both non-optimized and optimized models were documented to observe the performance of the models during the training and testing phase. At the final step, the model prediction results during the training and testing were evaluated by using mean absolute error (MAE), root mean square error (RMSE), and coefficient of determination (R 2 ) methods, which could be expressed by the following equations [ 11 , 27 , 31 ], …”

“…Machine learning (ML) is a subdivision of AI, has reinterpreted the world in diverse forecasting fields for the past two decades. The rapid advancement of graphics rendering and computer synchronization combined is the reason for the excessive growth of ML popularity than other prediction methods [ 11 ]. The ML models are capable of adaptive learning from the data, and it can improve themselves from subsequent training, trends, and pattern identification.…”