Ingredients, Recipes and recipe ratings. I would like to predict the rating based on combination of ingredients

Question

I would really appreciate some help on the first steps to my problem, suggestions of modeling techniques i could use or relevant research (i could not find any).

I have a list of ingredients (150 in total) and a list of recipes (100 or so) with data on which combination of ingredients are present in each recipe (on average 4-5 ingredients per recipe).

My output variables are 4 metrics of ratings of those recipes (taste, smell, look, and texture) for each of those 100 recipes from a scale of 1-10 for each.

My aim is to predict the 4 metrics based on a new combination of ingredients (picked from the list) that i specify.

What modeling technique should I look at to help solve this prediction problem?

Answer 1

Here are a few ideas to start with:

A simple model would be to use multiple linear regression (MLR), or a Random Forest

If you want to evaluate only based on if an ingredient is used, your input dataset could look like this:

Butter Flour Eggs ... Test Smell Look Texture 1 0 1 1 5 2 4

If you want to predict on how much of each ingredient (some is better, too much is not good) your dataset could look like this:

Butter Flour Eggs ... Test Smell Look Texture 1.5 3.0 2 1 5 2 4

You can get some other ideas from this website:

https://machinelearningmastery.com/multi-output-regression-models-with-python/

Answer 2

Random Forest can work good here since it is a decision tree. Before you call RandomForest, you will have to OneHotEncode your categorical variables e.g. Butter Flour Eggs ... because the Regressor or Classifier (whichever you fancy) cannot work with string and NaN values.

https://scikit-learn.org/stable/modules/generated/sklearn.ensemble.RandomForestRegressor.html