Data Study on the Culinary Footprint of World Cuisines
The skills the author demoed here can be learned through taking Data Science with Machine Learning bootcamp with NYC Data Science Academy.
We all love food; and we all talk about it a lot. There are myriad of data sets on the web for food reviews (Yelp), food patterns affecting health and lifestyle, census reports/surveys on food habits across age groups and the like.
However, in this space we often miss out to dig down and investigate the main point - composition and organization of food. Why does food taste the way it tastes? What is the influence of culture on our food habits? Is there any bias introduced by the cultures? How have cuisines evolved over time - are they becoming more healthy or unhealthy?
This blog post and its corresponding R shiny app is a sincere attempt to answer these questions. The app has extended benefits that can be leveraged for companies aimed at innovative food design and tweaking recipes for better nutrition and health.
Note: All the visualizations are interactive and have been built using Plotly
I used a private dataset that I had scarped during my undergrad. The data consists of :
- 1.5 million dishes/recipes
- 930 unique ingredients
- 6 world cuisines
- 24 regions
One of the most interesting and exciting module of my app is the
recipe generator. This features provides the user with an opportunity to discover and explore dishes from different cuisines/continents and regions and further refine the search by type of dish, ingredients to include and ingredients to exclude. 'Ingredients to exclude' is an unique feature which lets the user exclude ingredients that he/she dislikes or is allergic to.
The app takes a few seconds and presents the user with a comprehensive list of dishes from the rich culinary database that match his requirement.
The 930 unique ingredients have been classified into 20 categories namely Fruit, Vegetable, Spice, Pulses, Dairy etc. The app enables the user to see the distribution/prevalance of these ingredients over different cuisines, regions and subregions. For instance, the heatmap reveals a prefect prevelance (
1.0) of Spice category in Asia and Africa. This is in sync with our intuition - i.e. - every dish in the Asian and African cuisine contains alteast one spice ingredient. Similar trends can be observed for other cuisines and ingredient categories. Salient and subtle patterns in the use of different ingredients across the globe can be observed using this heatmap.
Furthermore, the user can also explore the
top n ingredients used in any cuisine, region or subregion.
The above graph indicates that all dishes in any cuisines follow a common pattern in their sizes. Most dishes contain an optimum number of ingredients - no dishes are overly simple or overly complex! Recipe sizes are bounded with a thin tail.
The average size of a dish is about 10 ingredients.
Try it yourself
Find the link to the app here to begin your investigations!