Estimation of sunflower production amount of Edirne province by machine learning methods

Sunflower, the most important oil plant in Turkey, in recent years , has decreased in production, and as a result of the increase in consumption, it has preferred to be imported to supply the need. However, there are also difficulties in the importation of sunflower due to epidemics, natural disasters and wars in the world. Especially in terms of satisfying the need for oils, in order to meet the need sunflower, it is essential to calculate the sunflower production accurately so that it can be imported in a appropriate quantities. A large amount of imports will cause excess stocks, and a small amount will cause the demand not to be supplied. In this study, it is aimed to use machine learning methods instead of quantitative estimation methods to calculate the amount of sunflower production. Because machine learning methods have features of easily determining the unknown or difficult to model relationships between data and making predictions with missing data. In the study, machine learning methods, Adaptive Network Based Fuzzy Inference System (ANFIS), Artificial Neural Networks (ANN), linear regression and decision tree have been used. The ANFIS method has used because of its has fast learning capacity and adaptability. ANN method is presented for comparison only. In order to increase the learning capacity of ANFIS, the input variables' membership functions have been determined via K- means algorithm. The estimated production in ANFIS, ANN, linear regression, decision tree models has been calculated in accord with the such as the cultivated area, humidity, temperature and the duration of insolation and the amount of rainfall. These models have been applied in Edirne region and ANFIS has given the least error (MSE). Thus, with the ANFIS model it is to be possible to estimate the amount of production in following years depending on the changing amount of inputs.