Development of strategic wheat crop prediction toolkit using machine learning algorithms to reduce food security risks (case study: alborz province)

Wheat as the main food in the country is of particular importance. Wheat is not only an important economic agricultural commodity in the world, but also known as a powerful lever in political and global relations. Therefore, the analysis and forecast of the production status of this product in the country has always been the focus of attention. The purpose of this study is to predict the amount of wheat yield (X) using artificial intelligence in the annual time scale in Alborz province. For this purpose, using annual cultivation and production data, wheat yield was investigated in six cities of Nazarabad, Savojbalagh, Karaj, Eshtehard, Fardis and Taleghan with a period of 40 years (1981-2020). After calculating the yield (ton per hectare) and forming an annual time series, four artificial intelligence methods including the best neighbor algorithm (KNN), backup vector (SVM), gene expression planning (GEP) and Bayesin Network (BN) were used and the wheat yield was predicted for the following year. Results indicated a more precision in yield prediction in the years with more production; According to the results of the BN, GEP, SVM and KNN model, the correlation coefficient between the observed and anticipated wheat yield values was 0.84, 0.89, 0.89 and 0.92, respectively. Explaining that Karaj and Taleghan cities have the highest and lowest wheat production respectively. The results showed that the KNN method had the best accuracy among the others, as the values of R, RMSE and MAE varied from 0.84 to 0.92, 0.21 to 0/24 and 0.11 to 0.18. Overall, by comparing the proposed methods, the KNN method had the highest and the BN method had the least accuracy to predict the amount of wheat yield in Alborz province. The results of this study can be very useful in providing and managing food security in areas under study.