Nowadays, nutrition and the food system are among the most important aspects
of daily life. In order to improve the performance of food recommendation systems, it would be beneficial to expand models that have a good understanding of the region and cooking style. This study attempts to predict the region in which a cooking recipe belongs based on the recipe and the textual and visual characteristics of the ingredients.
There is no comprehensive dataset available for this task. It is therefore necessary to collect the necessary data first. This study examines the problem of predicting the food region using textual and visual features extracted from the data. This is the first method that utilizes both textual and visual data to solve this problem. Additionally, over 600,000 images of food ingredients have been collected and labeled. Using the data collected on more than 100,000 cooking recipes in 26 regions, the model presented in this study has achieved an accuracy of 61.11%.