Using ANOVA and ANFİS Approaches in Modelling Agricultural Experiments

Abstract

Adaptive Neuro-Fuzzy Inference System (ANFIS) can analyze the factors and factor levels affecting the subject of interest in many branches such as technology, production, health, social and education, depending on the many rules it creates and with a very small experimental error (RMSE). and modelling. It is also applied in the field of agriculture, especially for the solution of problems such as agricultural field selection or technological product development. On the other hand, classical statistical methods are generally used in due diligence studies in a certain time period, such as product cultivation. Experimental design methods or in other words analysis of variance (ANOVA) methods come first among these methods. With the experiments modeled by ANOVA, the factors affecting the subject of interest and the levels of these factors are analyzed according to a single rule of the method used. Since the Root Mean Square Error (RMSE) of the model formed by the multiple rules of ANFIS versus the single rule of ANOVA is much smaller, it gives stronger results. Modeling agricultural products with ANFIS depending on time will support data mining studies in this field. In this study, first both ANOVA and ANFIS methods were briefly explained, and then the data of a due diligence study carried out in agriculture were modeled by both methods and similar findings were obtained. However, mostly the standard deviation (RMSE) values of ANFIS were found to be smaller than ANOVA. In addition, the relationships between ANFIS outputs and real measurements were examined.