Interactive Visualization
I learned from my family how to cook the cuisine from where they are from. It was important to us and represented a type of cultural identity. I was always really interested in food and have worked in several kitchens and restaurants. I wanted to look at cuisine types and find what ingredients are most common among them. This type of analysis lends itself to contrasting with economic and demographic data so some is provided here as well.
DATA - I found a good of amount of information to be available about recipes and cuisine types. Used here is a Kaggle dataset comprised of approximately 40,000 recipes from 20 different cuisines. To supplemnet to recipes dataset, I downloaded demographic and economic data grouped by country and joined this with the initial dataset. Some processing and cleaning was required and was done in Python.
METHOD - The best way to represent the distribution of recipes across cuisine types was spatially. I developed an interactive map in Python which I host on AWS. I created a R-Shiny version to include for comparison and it can be found by clicking the button below. The map displays cuisine specific data. The full dataset contained approximately 6,700 unique ingredients with a total count of all ingredients around 428,000. The data was processed to provide the top five most frequent ingredients per cuisine. The code can be found be clicking the black button below.
CONCLUSIONS - The analysis of ingredient types showed European cuisine’s focus on olive oil and garlic while cuisines of North East Asia contain more soy focused ingredients. Central Asian cuisines were found to use a lot of spice, while South American cuisines focused on salt and citrus. This analysis does not attempt to establish causality of health related issues such as obesity. The complexities and societal conditions that attribute to health are well beyond the scope here. The map serves to provide insight into cuisine types; insight that makes me want to no more about foods I do not have often.
The full dashboard can be found by clicking the image below
….And just because I think its cool, I included here a plot I made in Python you can click around on and drill down into years and albums.