Open AccessComparative Study of SVR, Regression and ANN Water Surface Forecasting for Smart Agriculture
Author(s) -
Arief Andy Soebroto,
Imam Cholissodin,
Destyana Ellingga Pratiwi,
Guruh Putra
Publication year - 2022
Publication title -
habitat
Language(s) - English
Resource type - Journals
eISSN - 2338-2007
pISSN - 0853-5167
DOI - 10.21776/ub.habitat.2022.033.1.9
Subject(s) - artificial neural network , support vector machine , mean squared error , linear regression , regression analysis , regression , backpropagation , proper linear model , statistics , artificial intelligence , machine learning , computer science , mathematics , bayesian multivariate linear regression
In the smart agriculture system based on green-based technology of artificial intelligence (AI), flooding can be predicted early by forecasting the water surface and good agricultural irrigation. The process of rising and falling of the water surface in a water basin area can be explained theoretically, but since there are many related variables and the complexity of dependencies between variables, the mathematical model is difficult to construct. Forecasting water surface in the field of irrigation needs too many variable parameters, such as cross-sectional area, depth, volume of rivers and so on. Based on patterns in each period, forecasting can be done using a statistical method and AI. This study uses the support vector regression (SVR) method, regression, multiple linear regression, and algorithm backpropagation, all compared to one another. The results of tests carried out between SVR and multiple linear regression show that SVR is superior. This can be seen from the result of the mean square error (MSE) obtained for each method. SVR 0.03 and for multiple linear regression, 0.05. The result is also supported by the best MSE result in the regression method, which is 0.338, and the best MSE value in artificial neural network (ANN), which is 0.428.