A Comparative Study Of Classification Methods Applied On Asian Cuisines Ingredients

Culinary similarity presents a significant part in the field of nutrition where it can be used to classify cuisines according to several factors such as culture,health,tourism or standard of living.However,people belong to different nationalities and regions have different dietary preferences.Thus,this differentiation of food recipes is characterized by the ingredients which are considered in this case as the atomic elements for each combination of foods(recipe).Discovering the relationship between cooking methods,food content and the geographic location of each country or region is fundamental for understanding the process of classifying certain cuisines based on their flavors or ingredients.Often,during a trip abroad or in a strange region,the human is always confronted with problems of food choices,because of not having sufficient information on the food to be consumed and especially when it is also confronted with a problem of maintaining a specific diet.Previously,some researchers have performed different computational methods to discover the similarity between food in a common environment for all recipes.But our approach focuses on analyzing a strong correlation between the geographic location of populations and their food preferences,and selects the best machine learning algorithm which is suitable for establishing the similarity between different combinations of ingredients called recipes that belong to particular geographical areas.So,in this study,we devised a novel classification approach using machine learning based on diet preferences of populations living in geographically remote areas.Further,finding the best machine learning algorithm to process this issue is very difficult due to the number of existing algorithms and their complexity in the terms of space and time for solving some problems.To achieve the objective of our study,first,we considered the sets of recipes which are defined in terms of Asian countries cuisines,i.e.China,Japan,India,Vietnam and Thailand.The food data for the related countries was downloaded from public website:www.bigoven.com.Then,to be sure of the outcome of this study,three algorithms such as:logistic regression,support vector machine,and neural networks were implemented and applied to the same set of data but with different parameters.Lastly,the results obtained showed that the support vector machine with a normalized polynomial kernel gave a better accuracy of classification and prediction of Asian cuisines compared to other methods mentioned in this study,which will contribute to food engineering investigations and help future researchers to overcome the major issues in this research area.