The idea

• I wanted to see how well you could visualize the difference between bread recipes by using data on their ingredients. I did this by, building an interactive search interface to explore recipes.

The Data

• I used the RecipeDB data (https://cosylab.iiitd.edu.in/recipedb). I cleaned the data limiting it to Bread recipes. I used Manifold Learning to reduce my 150 ingredient dimensions to a 2 dimensional representation. This is how I created the plot below. I used K-means clustering to organize the recipes into distinct groups.

The Build

• I performed my data analysis work using Python and the scikit-learn and pandas libraries. I used HTML, Javascript, and the d3 library to create the searchable table and graphic.

Takeaways

• Create the many interactions between the search bar, table and graphic was challenging but very rewarding. It was fun to play with interactive SVG components.

• I think my choice in Dimensionality Reduction worked reasonable well, however I would have liked to explore what it could have looking like with more munging.

• At the time of analysis, I did not have experience with other Clustering techniques. If I were to revisit this, I would try other algorithms such as Gaussian Mixture Models.