macromicro

Inspiration

It is my goal to eat healthier and in order to do that I have to know how different amounts of macronutrients and micronutrients relate to each other.

What it does

It shows the user a chart depicting the amount of nutrients of their choice, and it also provides a random forest regressor that takes two arrays in any size for the dependent micronutrient and independent macronutrient variables. This model predicts the amount of selected micronutrients from the amount of selected macronutrients.

How we built it

We took the USDA Nutrition dataset and performed basic analysis and exploration on different amounts of nutrients.

Challenges we ran into

Using the right machine learning model included general knowledge of both neural networks and classical machine learning models (XGB, Random Forest), as well as benefits and drawbacks in performance. Applying this new knowledge in its technical form is a hard undertaking to accomplish in just under ten hours while maintaining a proper sleep schedule.

Accomplishments that we're proud of

Planning of entire app and most notebook work pertaining to data analysis was completed by midday Saturday with several hours of sleep.

What we learned

Learned Streamlit, benefits and drawbacks of multiple classical machine learning models and how they compare to neural networks in terms of what kind of data is being used (size, type, complexity)

What's next for macromicro

functionality enhancements, functionality in general, diversity of models to choose from, better model performance metrics, opencv rgbd camera interface that uses CNN to detect food content and perform calculations instead.

Built With

• numpy
• pandas
• python
• scikit-learn
• streamlit
• xgboost