MyFoodPal

• 

  Welcome page for returning users

• 

  Previous diagnoses in a list

• 

  Welcome page for new users/ guests

• 

  User input page

Inspiration

We created a model to that predicts symptoms of various food groups from patient with irritable bowel syndrome (IBS). Using XGBoost. The model takes of concentration of food consumed as inputs and predicts their symptom score. From the generated score, we calculated the spearman correlation of each food group to find their contribution.

What it does

Our product, My FoodPal, is designed to help IBS (Irritable Bowel Syndrome) patients manage their symptoms by analyzing the impact of different food groups on their symptom scores. By taking into account the user's diet, My FoodPal calculates the IBS symptom score for each food group and provides personalized insights to help users better understand the relationship between their food choices and symptom severity. This information empowers patients to make informed decisions about their diet and effectively manage their IBS symptoms, ultimately improving their quality of life.

How we built it

We have created a postgres database hosted on aws servers. Made a backend with .net7 C# framework with all CRUD operations and APIs and added swagger documentation. Made a python flask api to get the prediction result. Made a react web app for the front end, which interacts with the API's defined in the swagger.

Challenges we ran into

The R squared between the test set and the prediction was significant and did not produce the desired outcome. With exploratory data analysis, we foresee tuning out in the future. We found that computing a correlation between target label and inputs, some food groups have a very high correlation and can be used as a new feature. Furthermore, using K-means algorithm, PCA and hierarchical clustering, we were able to find associations between features and finding features with the highest variance. The distribution of the training target label is also exponential, and scaling the targets to a log scale will also improve the accuracy.

Accomplishments that we're proud of

Learned various ways to perform EDA to create new features. Learned how to create a model and connect with a web application.

What we learned

Learned to collaborate and communicate as a new team with two members, Tahsin and Nicolas's, being in their first hacktahon.

What's next for MyFoodPal

We want to explore the other models and perform more extensive model selection and parameter tuning to reduce underfitting and overfitting.

Built With

• .net
• c#
• flask
• javascript
• postgresql
• python
• r
• react
• sql