Özet:

Purpose - The purpose of the research is to predict domestic food consumption classes in Turkey. Method - The data used in the study belongs to the Household Budget Survey collected by the Turkish Statistical Bureau. The data obtained from a total of 11,521 households for the year 2019 were used in the analysis of the study. The method of artificial nerve networks was used to determine the class of food consumption. In the artificial nervous networks method, 70% of the households used the model training, the remaining 30% was allocated for the test stage. The performance of the model established by the metrics obtained from this matrix is measured by the countermatrix used in the assessment of the problem of binary classification. 73% in the training stage. In the test phase 12, it has a 73.39% accuracy rate. The model’s sensitivity is 77.28% in the training stage and 80.39% in the test stage. The model’s accuracy is 71.22% in the test phase and 69.73% in the test phase. Metrics and metric comments are detailed in the findings section. When the methods calculated from the training and test phase results of the established model were studied, it was found that the system achieved correct and consistent results. Discussions - Except for the variables obtained from the studies studied in literature, the variables added to the model have been shown to increase the classification accuracy of the established artificial nervous networks model. The entire work of domestic food consumption in Turkey has been carried out using econometric models. It is believed that the econometric models that will be established with the added new variables will change the results of the previously achieved models. It is believed that using the method of artificial nerve networks, the work conducted in Turkey may contribute to literature because it has not been found a study in the literature aimed at predicting the non-home food consumption class.