Machine Learning 2

Case Analysis 2: Nutrition Machine Learning

The second case is about regression prediction in ML. The topic of the assignment is related to Food and Drug Administration (FDA), who along with many other things are also responsible for public awareness and education on food nutrition. FDA is the agency that makes food labels found on all packed and processed foods. FDA releases nutritional data which contains information on more than thirty nutrients found in over eight thousand food items. This is deterministic regression problem, which means a good solution exists and it is your task to find the best fitted model using various ML regression algorithms you will learn in the coming weeks.

You are given a curtailed form of the original Nutrition data. The original data contains 23 tables of various size, with the largest table containing 7 million rows. The original DB has information on more than 240,000 food items that are consumed in American diet. The tables were joined, stacked and arranged in structured form for this case. Only the relevant pieces of information and a small number of food items were kept for this assignment. The dataset contains a mere total of 8789 food items each described in a unique row. There are 43 columns in this dataset, where most of the columns represent the amount of nutrients within each food item, e.g. the first row represents salted butter that contains 717 calories, 0.85 g of protein etc. The last few columns show the standardized amount of the food item for which those nutrients were estimated by FDA.

Your task is to create regression models predicting Calories (column name: Energ_Kcal) using any number of predictors available to you.