# OBESEYE: Interpretable Diet Recommender for Obesity Management using Machine Learning and Explainable AI

**Authors:** Mrinmoy Roy, Srabonti Das, Anica Tasnim Protity  
**arXiv:** [2308.02796](https://arxiv.org/abs/2308.02796)  
**Date:** 2023-08-05  
**Summary page:** https://www.summarizepaper.com/en/paper/2308.02796v1/

## Key points

- Obesity is the leading cause of many non-communicable diseases
- Overeating and lack of physical activity contribute to obesity
- Healthy diet plans are important, especially for individuals with comorbidities
- Determining nutrient requirements for optimal health is challenging due to variations in physical and disease conditions
- The researchers developed a machine learning-based system called OBESEYE to predict personalized nutrient recommendations
- Various machine learning algorithms were used, including linear regression, support vector machine (SVM), decision tree, random forest, XGBoost, and LightGBM
- Fluid consumption and three major micronutrients (carbohydrates, protein, fat) were predicted using different algorithms
- High accuracy with low root mean square error (RMSE) was achieved in the predictions
- Chronic kidney disease (CKD) was the most important feature in overall fluid prediction according to SHAP summary plot
- Diabetes mellitus (DM) played a significant role in carbohydrate prediction
- Local explanations using LIME demonstrated how specific features influenced protein and fat predictions
- OBESEYE is unique in recommending personalized diets considering comorbidities and physical conditions
- The researchers hope that their system can help individuals manage obesity and improve overall health through healthy eating habits.

## Summary

In a study conducted by Roy et al., the researchers aimed to address the issue of obesity - the leading cause of many non-communicable diseases. They highlighted that overeating and lack of physical activity are major contributors to obesity. To combat this, they emphasized the importance of healthy diet plans, especially for individuals with comorbidities. However, determining the exact nutrient requirements for optimal health can be challenging due to variations in physical and disease conditions. To tackle this problem, the researchers proposed a novel machine learning-based system called OBESEYE. This system aims to predict personalized nutrient recommendations for maintaining good health. The team utilized various machine learning algorithms such as linear regression, support vector machine (SVM), decision tree, random forest, XGBoost, and LightGBM in their study. They focused on predicting fluid consumption as well as three major micronutrients: carbohydrates, protein, and fat. By using linear regression for fluid prediction and random forest for carbohydrate prediction, along with LightGBM for protein and fat prediction, the researchers achieved high accuracy with low root mean square error (RMSE). To provide further insights into their model's predictions and feature importance, they employed SHAP (Shapley Additive Explanations) for global interpretation. The SHAP summary plot revealed that chronic kidney disease (CKD) was the most important feature in overall fluid prediction while diabetes mellitus (DM) played a significant role in carbohydrate prediction. Additionally, local explanations were provided using LIME (Local Interpretable Model-Agnostic Explanations). These explanations demonstrated how specific features influenced protein and fat predictions. Overall, the researchers believe that their diet recommender system OBESEYE is unique in its ability to recommend personalized diets considering comorbidities and physical conditions. They hope that by promoting healthy eating habits through their system, individuals can effectively manage obesity and improve their overall health.