Food Recommendation System Using Machine Learning for Diabetic Patients

Aim of the Project

The aim of this project is to recommend proper foods for diabetic patients in the context of nutrition and food characteristics by developing a machine learning based intelligent mobile application.

Background

Diabetes is a huge problem. The number of diabetic patients increases each year by millions. And can you imagine only for the year 2021 the health expenditure for diabetes was more than 966 billion USD? Not only that. According to the research, 1 in 10 adults has diabetes and 1 death per every 5 seconds worldwide. Even in Sri Lanka, one in five adults has either diabetes or prediabetes. So how can we get rid of this? Unfortunately, according to medical professionals, there is no cure yet.

The only thing, we can do is to control diabetes, especially by avoiding unhealthy foods. So how can we scientifically ensure whether a particular food is healthy or not for diabetes? Well, this is where the Glycemic Index comes into play. The Glycemic Index is a rating system for foods and foods with a low Glycemic index are recommended for diabetic patients.

But have you ever seen the Glycemic Index printed on packaged foods? Most probably no right? In fact, the glycemic index can only be determined inside a laboratory.

Diabeedoc which is an intelligent mobile application. Using Diabeedoc you can find the suitability of a particular food for diabetes in no time. We have deployed our GI based machine-learning models to the Azure cloud. Diabeedoc uses those ML models with the Openfoodfacts database to predict particular food is suitable or not. If it is not suitable Diabeedoc is capable of suggesting better foods using machine learning based substitution system.

Expected Outcomes

• Suggest whether a particular food is healthy or not for diabetic patients
• Train a machine learning model to suggest foods with an approximate amount of nutritional values (specifically for unhealthy foods)
• Predict whether a newly introduced food is suitable for diabetic patients and if the food is not suitable, suggest healthy foods with an approximate amount of nutritional values
• Develop a user friendly interface to access the recommendation system