Dietary Recommendation System Based on Machine Learning

G. Lucy, K. Meghana, Damagatla Sowmya, G. Rajeswari and K. Suma Latha

Department of Computer Science and Engneering, Ravindra College of Engineering for Women,

Kurnool, Andhra Pradesh, India

Keywords: Diet Recommendation, Machine Learning, Decision Tree Classifier, Personalized Nutrition, Health

Monitoring, AI in Healthcare, Dietary Planning, Chronic Disease Management, Nutritional Guidance,

Data-Driven Dieting.

Abstract: In order to make nutrition recommendations based on machine learning, this research employs a Decision

Tree Classifier. Dietary restrictions, physical activity levels, age, gender, BMI, disease type, severity, and

training data are all inputs into the model. The system accurately predicts outcomes, encodes categorical

factors, and partitions the dataset for testing and training. Diet recommendations tailored to the user's input

can be made in real-time by saving the trained model. For optimal health, the model suggests tailored meal

patterns that include Low-Carb, Low-Sodium, and Balanced foods.

1 INTRODUCTION

In today's fast-paced world, it is crucial for well-being

to eat healthily. It is rare for traditional diet regimens

to account for a person's unique health status, food

restrictions, and way of life. Poor diet adherence and

the risk of chronic diseases are outcomes of many

people's struggles to choose meals that fulfil their

health demands. More than ever before, there is a

demand for tailored, data-driven dietary advice.

Better patient outcomes can be achieved by data-

driven healthcare decision-making made possible by

machine learning. By analysing large amounts of

health data, machine learning algorithms can spot

patterns and provide reliable forecasts. Incorporating

body mass index (BMI), disease kind and severity,

degree of physical activity, and dietary preferences

into personalised nutrition plans might be a machine

learning-based approach. People are better able to

achieve their health goals and benefit from nutritional

interventions as a result.

A Decision Tree Classifier is employed in the

proposed method to provide dietary recommendations

according to health criteria. A huge dataset including

demographic information (gender, age, weight, blood

pressure, glucose, and food restrictions) is used to

train the algorithm. Following training, the model

might recommend a Balanced, Low-Carb, or Low-

Sodium diet depending on the patient's condition.

Quickly and accurately, this automated diet planner

provides guidance.

An important advantage of this technology is its

ability to quickly assess user input and provide dietary

recommendations. Customers receive practical and

attainable dietary guidance from the system in the

form of personalised meal plans based on the

anticipated diet category. Over time, the model

becomes better at delivering accurate and useful

suggestions as it learns from new data.

Fast, quick, and tailored healthy eating is now a

reality with this technology that suggests diets based

on machine learning. It improves people's quality of

life by using modern data science to better manage

their health and food. By offering individualised meal

plans grounded in scientific research, this method

contributes to the promotion of long-term health.

2 LITERATURE SURVEY

2.1 Personalized Diet Recommendation

System Using Machine Learning:

Individuals can get health and nutrition advice from

the "Personalised Diet Recommendation System"

using machine learning. Using user-supplied data, this

project constructs an ML model to provide dietary and

health-related suggestions tailored to each individual.

278

Lucy, G., Meghana, K., Sowmya, D., Rajeswari, G. and Latha, K. S.

Dietary Recommendation System Based on Machine Learning.

DOI: 10.5220/0013911800004919

Paper published under CC license (CC BY-NC-ND 4.0)

In Proceedings of the 1st International Conference on Research and Development in Information, Communication, and Computing Technologies (ICRDICCT‘25 2025) - Volume 4, pages

278-282

ISBN: 978-989-758-777-1

The model takes into account factors such as age,

gender, dietary goals, activity intensity, and weight

loss targets. The next step is for it to say if the user is

slim or not. After doing the necessary analysis, the

system will provide meal plans for the day. You may

find nutritional information and cooking directions for

individual meals in these programs. It is advised that

at each meal, favoured foods be selected from a

variety of categories. The system takes a person's

dietary consumption into account and displays the

results graphically in a pie chart. User preferences,

body mass index (BMI), and physical characteristics

inform our project's output designs.

2.2 AI Nutrition Recommendation

Using a Deep Generative Model and

ChatGPT:

Advanced AI nutrition systems have recently

emerged, thanks to recent advancements in AI, to

enhance tailored dietary guidance and overall health

and wellness. There are concerns regarding the

accuracy and reliability of AI systems' nutritional

recommendations because to the absence of

professional norms. A state-of-the-art AI nutrition

advising system that adheres to dietary rules is

developed using new complex loss functions in

conjunction with the speed and explainability of deep

generative networks. Make accurate predictions about

users' anthropometrics and medical conditions in

descriptive latent space using a variational

autoencoder and an optimiser to alter meal portions

depending on energy demands, the system provides

customised, healthy, and highly accurate weekly meal

plans. The proposed method has the potential to

increase meal diversity, accuracy, and generalisability

using ChatGPT's unparalleled library of meals from

many cuisines. The suggested diet recommendation

method can be readily incorporated into upcoming

diet recommendation systems and produces weekly

meal plans that fulfil users' dietary and energy

requirements, according to extensive trials conducted

on 84000 daily meal plans for virtual users and 3000

real users with 7,000 daily meal plans.

2.3 A Food Recommender System

Considering Nutritional

Information and User Preferences

Health problems that do not spread from person to

person are on the rise, according to the World Health

Organisation. These include cancer, diabetes, and

premature heart disease. Poor dietary habits are

associated with several diseases. Individuals' unique

physical, physiological, and personal traits inform a

novel area of research known as "personalised

nutrition," which provides dietary recommendations.

Specifically, by integrating user data with nutritional

information, several recent research have created

computer models for customised meal selection.

Unlike previous attempts, this research presents a

global framework for recommendation of daily meal

plans that handles information on preferences and

nutrition at the same time. The idea uses AHPSort, a

method for multi-criteria decision analysis, to screen

out foods that won't work for the people using the

concept right now. A daily meal plan that takes into

account the user's preferences, eating habits, and

nutritional requirements is generated through an

optimization-based stage. Using a case study, the

recommender system is tested.

2.4 AI-Driven Nutritional Assessment

Improving Diets with Machine

Learning and Deep Learning for

Food Image Classification:

For optimal health and to stay away from food-borne

illnesses, a well-balanced diet is a must. In order to

address this critical public health concern, we employ

DL techniques and ML to categorise food photos and

anticipate important attributes. For effective good and

bad food product classification, our approach employs

a sophisticated hybrid model that mixes a deep

learning CNN with a SVM. Using the SVM classifier

for classification, the CNN-based method streamlines

feature extraction. Using a tailored dataset, we

evaluated our approach. With a 97% and 94%

accuracy rate, respectively, our hybrid model

outperforms the CNN model in the experiments.

Improvements in recollection, f1-score, and accuracy

are also seen.

2.5 Intelligent Personalized Nutrition

Guidance System Using IoT and

Machine Learning Algorithm

There has been a global uptick in dietary issues.

Problems like diabetes, obesity, and weight gain can

stem from a poorly balanced diet. The system is able

to assess food images in novel ways to suggest better

eating habits thanks to the integration of image

processing. To get useful information out of food

photo data, we combine machine learning, the internet

of things (IoT), and image processing. Images of food

taken using smartphones and other specialist cameras

are uploaded to the cloud for analysis. This study

Dietary Recommendation System Based on Machine Learning

279

proposes a novel approach to building a system that

gives individualised dietary recommendations by

utilising SVM and IoT technology. SVM searches for

correlations, trends, and dietary requirements in this

data. All eating habits are saved in our database on the

cloud. Calories and nutrients may be calculated using

this method, which makes use of image processing

and segmentation. The nutritional value, serving size,

and food type labels provide the algorithm with a

wealth of dietary data from which to learn. This

trained SVM model is utilised by the system to

evaluate the nutritional requirements, deficiencies,

and personalised dietary goals of users.

3 METHODOLOGY

3.1 Proposed System

The proposed system utilizes machine learning,

specifically a Decision Tree Classifier, to generate

personalized diet recommendations based on an

individual's health attributes. By analyzing various

parameters such as age, BMI, disease type, severity,

physical activity level, and dietary restrictions, the

system predicts the most suitable dietary plan,

including Low-Carb, Low-Sodium, or Balanced

Diets. The model is trained on a diverse dataset to

ensure accurate classification and effective meal

suggestions tailored to the user’s specific health needs.

To enhance efficiency, the system automates data

processing, encodes categorical values, and performs

real-time analysis to deliver instant dietary

recommendations. Users simply input their health

information, and the model evaluates their data to

suggest an appropriate meal plan. Unlike traditional

methods that require manual consultation, this AI-

driven approach provides quick, reliable, and

scientifically backed dietary advice.

Among the many benefits of this approach is its

ability to generate customized meal plans, ensuring

that individuals receive practical and actionable

dietary guidance. The automation reduces human

intervention, making it a more accessible and scalable

solution for personalized nutrition planning. By

continuously learning from new health data, the

system refines its recommendations over time, leading

to improved adherence and better health outcomes for

users.

3.2 System Architecture

The architecture of the suggested AI-driven diet

recommendation system consists of multiple stages,

ensuring efficient processing and accurate dietary

predictions (figure 1). Initially, user input is collected,

including health attributes such as age, BMI, disease

type, severity, physical activity level, and dietary

restrictions. This data undergoes preprocessing, where

missing values are handled, categorical variables are

encoded, and numerical features are normalized for

optimal model performance.

The Decision Tree Classifier is then applied to

analyze the processed data, identifying patterns and

relationships between health parameters and dietary

requirements. Based on the classification, the model

predicts an appropriate diet plan, such as Low-Carb,

Low-Sodium, or Balanced Diets. The system then

generates personalized meal plans, ensuring users

receive actionable dietary guidance tailored to their

specific health conditions.

Figure 1: Proposed Architecture.

3.3 Modules

3.3.1 User

In this application user is a module, should register

with the application then only he can login into his/her

account. After successful login he/she can able to enter

the input data and can view recommended diet and diet

plan and logout.

3.3.2 Admin

Here admin is a module can login directly with his

predefined username and password after successful

and applying preprocess to get all categorical columns

and applying label encoding then split data into

ICRDICCT‘25 2025 - INTERNATIONAL CONFERENCE ON RESEARCH AND DEVELOPMENT IN INFORMATION,

COMMUNICATION, AND COMPUTING TECHNOLOGIES

280

training and testing, building decision tree model and

save model and logout.

3.4 Algorithms - Decision Tree

Classifier

The Decision Tree Classifier is used in this project to

predict personalized diet recommendations based on

various health parameters. It operates by recursively

splitting the dataset into nodes, where each node

represents a decision based on a specific feature (such

as BMI, disease type, or activity level). The tree

structure allows the model to classify users into

different diet categories, such as Low-Carb, Low-

Sodium, or Balanced Diets, based on their unique

health profiles.

During training, the algorithm processes a dataset

containing age, gender, BMI, disease severity,

physical activity level, and dietary restrictions as input

features. The Decision Tree Classifier uses entropy or

Gini index to determine the best feature for splitting

the data at each step. A final dietary recommendation

is given when the tree reaches a node in the leaf

canopy.

Once trained, the model can quickly classify new

user inputs by traversing the decision tree and

selecting the most appropriate diet plan. For example,

if a user has high BMI, diabetes, and low physical

activity, the model may classify them under the Low-

Carb Diet category. Similarly, a user with

hypertension and moderate activity levels may be

recommended a Low-Sodium Diet.

The key advantage of using a Decision Tree

Classifier is its ability to provide clear, interpretable

recommendations while efficiently handling

categorical and numerical data. Additionally, the

model can be continuously updated with new dietary

data, improving accuracy over time and ensuring users

receive the most relevant nutritional guidance.

4 EXPERIMENTAL RESULTS

Accuracy: How well a test can differentiate between

healthy and sick individuals is a good indicator of its

reliability. Calculate the proportion of true positives

and negatives to evaluate the reliability of the test.

After doing the maths:

Accuracy = TP + TN /(TP + TN + FP + FN) (1)

Precision: The accuracy rate of a classification or

number of positive cases is known as precision.

Accuracy is determined by applying the following

formula:

Precision = TP/(TP + FP) (2)

Recall: The recall of a model is a measure of its

capacity to identify all occurrences of a relevant

machine learning class. A model's ability to detect

class instances is shown by percent of correctly

anticipated positive observations relative to total

positives.

Recall = TP/TP+FN (3)

F1-Score: A high F1 score indicates that a machine

learning model is accurate. Improving model

accuracy by integrating recall and precision. How

often a model gets a dataset prediction right is

measured by the accuracy statistic.

F1 Score = 2 / ( (1 / Precision) + (1 / Recall) ) (4)

F1 Score = (2 × Precision × Recall) / (Precision +

Recall) (5)

Upload dataset, View dataset,

Enter input data and

results pages are shown from figures 2-5 respectively.

Figure 2: Upload dataset.

Figure 3: View dataset.

Dietary Recommendation System Based on Machine Learning

281

Figure 4: Enter input data.

Figure 5: Results.

5 CONCLUSIONS

Customised diet plans could be radically altered by

AI-powered meal recommendation systems. Dietary

recommendations are generated by these systems

using decision trees, which take into account health

measures, dietary preferences, and medical issues.

When compared to more conventional methods, AI-

generated recommendations are superior for

managing chronic diseases like diabetes and

hypertension, as well as for ensuring patient

adherence. With the advancement of AI and real-time

data monitoring, these systems will be able to better

encourage healthy eating habits and provide improved

health outcomes.

6 FUTURE SCOPE

The diet recommendation system can be enhanced by

integrating wearable devices to track real-time health

metrics like heart rate, calorie expenditure, and

physical activity. Future developments may

incorporate deep learning models for improved

accuracy in dietary predictions and personalized

recipe generation based on user preferences, allergies,

and cultural habits. Adding voice assistant support can

make the system more interactive and user-friendly.

AI-powered food recognition technology can help

users monitor their daily food intake automatically.

Additionally, deploying the system as a mobile or

cloud-based application will improve accessibility,

allowing users to receive real-time diet

recommendations anytime, anywhere. These

advancements will make AI-driven nutrition planning

more precise, adaptive, and effective in promoting

long-term health and wellness.

REFERENCES

R. Kumar, S. Srivastava, & P. Tiwari, "Personalized Diet

Recommendation System Using Machine Learning,"

International Journal of Computer Science and

Information Security (IJCSIS), 2021.

P. Gupta, A. Sharma, & S. Verma, "A Machine Learning

Approach for Personalized Nutrition

Recommendation," Journal of Artificial Intelligence

Research, 2022.

A. Singh & P. Agarwal, "Dietary Pattern Prediction Using

Decision Tree Algorithm," International Conference on

Computational Intelligence and Data Science

(ICCIDS), 2020.

N. Ahmed, M. F. Khan, & A. Beg, "AI-based Diet Planning

for Diabetic Patients," IEEE Transactions on

Healthcare Informatics, 2019.

T. J. Theis & R. C. White, "Nutritional Recommendation

System Based on Deep Learning," Journal of Medical

Informatics and Decision Making, 2023.

Ian Goodfellow, Yoshua Bengio, & Aaron Courville,

"Deep Learning," MIT Press, 2016.

Christopher M. Bishop, "Pattern Recognition and Machine

Learning," Springer, 2006.

Trevor Hastie, Robert Tibshirani, & Jerome Friedman,

"The Elements of Statistical Learning: Data Mining,

Inference, and Prediction," Springer, 2017.

Géron Aurélien, "Hands-On Machine Learning with Scikit-

Learn, Keras, and TensorFlow," O'Reilly Media, 2019.

Tom Mitchell, "Machine Learning," McGraw-Hill, 1997.

Scikit-Learn Documentation, "Decision Trees in Machine

Learning," Available at: https://scikit-learn.org

TensorFlow Blog, "Building AI Models for Health and

Nutrition," Available at: https://blog.tensorflow.org

Kaggle, "Diet Recommendation System Dataset,"

Available at: https://www.kaggle.com

Harvard Health Publishing, "The Role of AI in Personalized

Nutrition," Available at:https://www.health.harvard.ed

Food and Nutrition Research Journal, "Dietary Data

Analysis with Machine Learning," Available at:

https://www.foodandnutritionjournal.com

ICRDICCT‘25 2025 - INTERNATIONAL CONFERENCE ON RESEARCH AND DEVELOPMENT IN INFORMATION,

COMMUNICATION, AND COMPUTING TECHNOLOGIES

282