Dogo Rangsang Research Journal UGC Care Group I Journal

ISSN : 2347-7180 Vol-13, Issue-4, April 2023

DIET RECOMMENDATION SYSTEM USING MACHINE LEARNING

[email protected], [email protected] ,
[email protected] , [email protected]*

Reema Golagana, V. Sravani, T. Mohan Reddy, IV year B.Tech students, Dept. of

Computer Science Engineering, Raghu Engineering College, Visakhapatnam, Andhra Pradesh, India
CH Kavitha Assistant professor, Dept. of Computer Science Engineering, Raghu Engineering
College, Visakhapatnam, Andhra Pradesh, India

ABSTRACT

In today’s culture, it is not easy to recommend a diet right away. Nowadays, people suffer from a range
of diseases and chronic illnesses. And for most cases, an unhealthy or improper diet is the root cause
of these ailments. On average, a person requires 2000 calories a day, however, the precise number of
calories consumed depends on a person's physical characteristics, including BMI, age, gender,
cholesterol, blood pressure, and other factors. In this paper, we propose a diet recommendation system
based on the user’s physics details and ailments. This research includes different machine learning and
deep learning techniques. The system learns from its training data, which includes daily calorie intake
and food consumption patterns, to predict the recommended meal plan for a user. We evaluated the
system on a dataset, and the results show that the proposed model outperforms the existing diet
recommendation systems. The proposed system predicts the meal plan based on the user's dietary
nutritional requirements. It could achieve a personalized diet plan for individuals and help individuals
lead healthier lives.

Keywords: diet, calories, BMI, diet recommendation, machine learning, deep learning

1. INTRODUCTION
People today struggle with a wide range of health issues, including mental health issues, poor diet, and
fitness issues. Numerous studies have demonstrated that a poor diet is a major contributor to a wide
range of health issues. A WHO study found that an inadequate and unbalanced diet is responsible for
approximately 9% of deaths from heart attacks, 11% of deaths from heart disease (Ischemia), and 14%
of deaths from gastrointestinal cancer worldwide, reportedly causing. In addition, approximately 250
million children suffer from vitamin A deficiency, 200 million from iron deficiency (anemia), and 700
million from iodine deficiency[7]. A diet recommendation system is a tool that assists people in making
healthy food choices. Diet recommendation systems analyze diet preferences, nutritional needs, and
other dietary factors to provide personalized recommendations for healthier eating. By the user’s age,
height, and weight, BMI calculates weight status categories which include underweight, healthy,
overweight, and obese. This project will help users with daily diet recommendations along with BMI
range, healthy food choices, eating behavior, and health problems. Seasonal foods, user-intreated
foods, plant foods, and animal products are all included in the grocery data. This project will assist
users with daily dietary guidelines, BMI ranges, appropriate food selection, eating behaviors, health
issues, and behavior modification. [8,9]

Page | 118 DOI:10.36893.DRSR.2023.V13I04.118-125 Copyright @ 2023 Authors

Dogo Rangsang Research Journal UGC Care Group I Journal
ISSN : 2347-7180 Vol-13, Issue-4, April 2023
There are several parts to the paper. An introduction to the use of machine learning techniques for diet
recommendations is given in the first part. A review of related literature is presented in the second
part. The third part of the study's methodology includes a description of the dataset, data pre-
processing, algorithms used, and various techniques like Random forest, K-means, and LSTM
algorithms. The findings and their analysis are presented in the fourth section. The conclusion, which
concludes the essay, also goes over its potential scope.

PROBLEM STATEMENT
The alarmingly high rate of consumption of fast food has resulted in the intake of unhealthy food.
Obesity, diabetes, high blood pressure, and other health issues are all a result of this. Thus it has
become extremely fundamental for individuals to have a decent adjusted to a healthy, nutritional, and
balanced diet. However, not everyone in today‘s society has the time or resources to afford a personal
dietitian or a nutritionist to take care of their health and provide a balanced diet plan based on their
unique circumstances. In this report, we discussed a person's poor nutritional eating patterns and made
an effort to provide them with a solution that would enable them to lead healthy lifestyles.

2. LITERATURE REVIEW
This literature review examines papers and studies that investigated the application of machine
learning algorithms to food recommendation systems for diets.
Celestine Iwendi et al. (2020) [1] investigate the data collection potential of their system. Machine
and deep learning algorithms such as Naive Bayes, Logistic Regression, Multilayer Perceptron (MLP),
Gated Recurrent Units (GRU), Recurrent Neural Networks (RNN), and Long Short-Term Memory
(LSTM) are the goals of this research framework and how they apply to IoMT data. For consideration
in the clinical dataset, 30 people's information with 13 highlights of different illnesses, as well as 1000
things, were accumulated through the internet and hospitals. There are eight features in the product
area. Before applying deep learning and machine learning-based techniques, the characteristics of this
IoMT data were examined and further encoded.
Thi Ngoc Trang Tran et al. (2021) [2] show that their method can be used to present a comprehensive
review of healthcare recommender system research: Besides, our exploration recognizes from past
important outlines concentrates that it gives knowledge for suggested circumstances and approaches.
Dietary recommendations, drug ideas, health status forecasts, service recommendations, and
recommendations from healthcare professionals are all included in this kind of proposal. They also
give students examples from real-world situations to help them fully comprehend recommendation
systems.
Gao et al. (2017) [3] introduced a computational framework for a personalized diet recommendation
system. The system used Bayesian personalized ranking along with matrix factorization to learn user
preferences from a wide range of data. The results showed that the approach led to better-personalized
recommendation performance than conventional collaborative filtering techniques.
Butti Gouthami and Malige Gangappa (2020) [4] The USDA nutrition dataset will be used to determine
the user's suggested diet. A set of grocery shop information that takes the user's preferred food intake
into account. The USDA database contains nutritional data for every food item. A USDA ID is used
as the baseline value for the input values for every 100 grams. Since they will ultimately be used to
estimate the suggested diet, the data required to compute BMI (body mass index) must be provided.

Page | 119 DOI:10.36893.DRSR.2023.V13I04.118-125 Copyright @ 2023 Authors

Dogo Rangsang Research Journal UGC Care Group I Journal
ISSN : 2347-7180 Vol-13, Issue-4, April 2023
The second input utilized to determine the user's suggested diet is their daily food intake. The diet
recommendation is initially created using the food that was eaten that day, and the input nutrient dataset
is sorted by a nutritional deficit.
Yera et al. [5] proposed a food recommendation system that takes user preferences and nutritional data
into account. The user's advised meal plan considered their preferences. This tool maintains both user
preferences and nutritional information
Iigo Orue Saiz et al. [6] made an effort to find the research that has been done and the recommendation
systems that have been used in the major databases for this purpose over the past five years. The results
lead to the following conclusion: Previous works focus more attention on the recommendation
system—typically collaborative filtering—than they do on the data or sample description; It is
unknown which indices are used to calculate calories or nutrients. In this way, to be viable, it is basic
to work with open information or all-around depicted information, which permits the experience to be
rehashed by different gatherings, or at any rate to be comparable.

3. METHODOLOGY
I. DATASETS
The recommended diet for the user will be determined using the USDA nutrition information. Every
food item's nutrition information is maintained in the USDA database. [5,19]
In this diet recommendation system, to calculate the recommended diet, the nutrient dataset is first
sorted based on the BMI value, and deficit nutrition is determined using the food consumed that day.
The recommender provides dietary advice. The samples of datasets used are in below figures (1) and
(2).

Figure 1: Sample food dataset

Page | 120 DOI:10.36893.DRSR.2023.V13I04.118-125 Copyright @ 2023 Authors

Dogo Rangsang Research Journal UGC Care Group I Journal
ISSN : 2347-7180 Vol-13, Issue-4, April 2023

Figure 2: Sample nutrient dataset

II. DATA PREPROCESSING

Data preprocessing is a step taken to prepare data for modeling with LSTM (Long Short-Term
Memory) networks. Preprocessing helps in transforming raw data into a format that is suitable for
analysis, reduces noise, and enhances the quality of the data. Cleaning the data and making it suitable
for a machine-learning model, which also improves a model's accuracy and effectiveness is a required
task. Feature Engineering is the process of enhancing or selecting significant features from the dataset.
The process involves identifying the relevant variables, removing unwanted features, and creating
derived features.
Train-Test Split is partitioning the available dataset into training and testing sets is crucial in evaluating
the model’s performance. LSTM can memorize the training data too well and result in overfitting.
Therefore, the model needs to be validated using an independent test set.
III. RANDOM FOREST
Random Forest is a popular machine-learning algorithm used in recommendation systems. It is
particularly effective for solving high-dimensional and non-linear problems. Random Forest is a type
of supervised learning algorithm that belongs to the family of decision trees. Unlike traditional decision
trees that work with a single tree, Random Forest combines multiple decision trees to create a more
accurate and robust prediction model. In a recommendation system, Random Forest can be used to
create a model that predicts the user's preferences based on their past behavior and the behavior of
similar users. The algorithm can analyze large datasets and extract patterns and insights that can be
used to make personalized recommendations.[7,18]
IV. K-MEANS
K-means clustering is a popular unsupervised machine learning algorithm in various applications,
including diet recommendation systems. If we have a dataset of users where each user has provided
information about their food choices, allergies, and dietary restrictions, we can apply k means
clustering to group similar users together. This can help in generating personalized diet plans, recipes,
and food recommendations for each cluster. K-means clustering works by assigning each user to the
nearest cluster based on their similarity.[18]
V. LONG SHORT-TERM MEMORY(LSTM)
An improved LSTM (long short-term memory) that accurately and precisely meets patient needs and
overcomes this problem in the existing model. LSTMs use a combination of an input gate, an output
gate, and a forget gate to selectively retain or discard information at each time step. This allows the

Page | 121 DOI:10.36893.DRSR.2023.V13I04.118-125 Copyright @ 2023 Authors

Dogo Rangsang Research Journal UGC Care Group I Journal
ISSN : 2347-7180 Vol-13, Issue-4, April 2023
network to maintain long-term dependencies in the input sequence and avoid the problem of
information loss over time.

Figure 3: LSTM architecture[15]

 Forget Gate: The forget gate determines what information must be remembered and what can
be forgotten. (Ct -1).
 Input Gate: The input gate is provided with new informational sets, which it learns and stores
in memory (Ct).
 Output Gate: The output gate gives information to the next hidden state
LSTM model is trained on a dataset of food logs, where each log consists of a list of foods consumed
in order. The LSTM could learn to predict what the user is likely to eat next based on their past eating
habits and could generate recommendations for healthier alternatives. [13,15,16,17]

4. RESULT
This diet recommendation system can help individuals achieve their health goals by providing
personalized dietary recommendations based on their age, weight, height, and preference, such as veg
or non-veg. The recommender suggests 3 types of diets – weight loss, weight gain, and healthy. diet.
After entering the personal details in the input dialog box, the BMI is calculated and it will show based
on the BMI if it’s underweight, overweight, or healthy.
The Root Mean Squared Error (RMSE) is one of the two primary performance metrics. It determines
the typical difference between actual values and a model’s predictions. It provides an estimate of the
model’s accuracy (predictive power) for the desired value.

Figure 5: RMSE formula [14]

Where,
Ŷ1, Ŷ 2, Ŷ3, ..., Ŷn - predicted values
Y1,Y2, Y3, ….., Yn - observed values

Page | 122 DOI:10.36893.DRSR.2023.V13I04.118-125 Copyright @ 2023 Authors

Dogo Rangsang Research Journal UGC Care Group I Journal
ISSN : 2347-7180 Vol-13, Issue-4, April 2023
‘n’ - number of observations

These error rates based on RMSE (Root Mean Square Error) show the LSTM has a slightly higher
accuracy than the Random Forest algorithm as shown in table (1) and graph (1) below.

Machine Learning Error rate

Models
Random Forest 0.3707

LSTM 0.3658

Table 1: Error rates

The graphical representation of error rates is in the below graph (1) :

0.372

0.371 0.3707

0.37

0.369

0.368

0.367

0.366 0.3658

0.365

0.364

0.363

Random Forest LSTM

Graph 1: Comparison Graph (Error Rates).

5. CONCLUSION AND FUTURE SCOPE

Machine learning algorithms have shown great promise in personalized diet recommendation systems.
Motivating people to consume a healthy diet is the aim of nutrition education. Dietary interactions that
are crucial for developing dietary guidelines are given special consideration. a health-based medical
dataset that uses features like age, gender, weight, and height to automatically identify which foods

Page | 123 DOI:10.36893.DRSR.2023.V13I04.118-125 Copyright @ 2023 Authors

Dogo Rangsang Research Journal UGC Care Group I Journal
ISSN : 2347-7180 Vol-13, Issue-4, April 2023
should be given to which patients based on their conditions. The machine learning and deep learning
algorithms used in this study framework include Random Forest, K-Means, and Long Short-Term
Memory (LSTM). The optimum eating strategy that yields positive effects are found in all of the
models presented in this study. The development of personalized food advice systems could be
furthered by investigating new important data sources and thinking about how to incorporate important
variables like physical activity, lifestyle, and disease history.
REFERENCES
[1] Celestine Iwendi, Suleman Khan, Joseph Henry Anajemba, Ali Kashif Bashir, And Fazal Noor,
“Realizing an Efficient IoMT-Assisted Patient Diet Recommendation System Through Machine
Learning Model”, IEEE Access, vol. 8,2020, pp. 28462-28474, doi: 10.1109/ACCESS.2020.296853.
[2]Thi Ngoc Trang Tran, Alexander Felfernig, Christoph Trattner and Andreas Holzinger,
“Recommender systems in the healthcare domain: state-of-the-art and research issues”, Journal of
Intelligent Information Systems, vol. 57, 2020, pp. 171–201, doi: 10.1007/s10844-020- 00633-6.
[3] Jun Gao, Ninghao Liu, Mark Lawley, and Xia Hu, “An Interpretable Classification Framework for
Information Extraction from Online Healthcare Forums”, Journal of Healthcare Engineering, Cui Tao,
vol. 2017, Article ID 2460174, https://doi.org/10.1155/2017/2460174.
[4] Butti Gouthami and Malige Gangappa, “NUTRITION DIET RECOMMENDATION SYSTEM
USING USER’S INTEREST”, International Journal of Advanced Research in Engineering and
Technology (IJARET), vol. 11, 2020, pp. 2910-2919, doi: 10.34218/IJARET.11.12.2020.272
[5] Rachel Yera Toledo, Ahmad A Alzahrani And Luis Martinez, “A Food Recommender System
Considering Nutritional Information And User Preference”, IEEE Access, vol.7,2019, pp. 96695-
96711, doi: 10.1109/ACCESS.2019.2929413.
[6]Inigo Orue-Saiz, Miguel Kazarez, And Amaia Mendez-Zorrilla, “Systematic Review Of Nutritional
Recommendation Systems”, Applied Science, vol. 11, no. 24: 12069, 2021, doi:
https://doi.org/10.3390/app112412069.
[7] Shubham Singh Kardam, Pinky Yadav, Raj Thakkar, Prof Anand Ingle, “Website on Diet
Recommendation Using Machine Learning”, International Research Journal of Engineering and
Technology (IRJET), vol.8, 2021, p-ISSN: 2395-0072.
[8] Salliah Shafi Bhat and Gufran Ahmad Ansari, “Predictions of Diabetes and Diet Recommendation
System for Diabetic Patients using Machine Learning Techniques”, International Conference for
Emerging Technology (INCET), 2021, Belagavi, India, pp. 1-5, doi:
10.1109/INCET51464.2021.9456365.
[9] S. V. N. Sreenivasu, Sakshi Gupta, Ghanshyam Vatsa And Anurag Shrivastava, “Carbohydrate
Recommendation for Type-1 Diabetics Patient Using Machine Learning”, 5th International Conference
on Contemporary Computing And Informatics, UP, 2022, pp. 600-604, doi:
10.1109/IC3I56241.2022.10072919.
[10] Aryan Jalali, Kumar P. Arjun Manoj, N.Amulya, Ayesha Siddiqui, Alok Kumar Singh, S.M. Hari
Krishna, “Mobile Application for Personalized Food Recommendation”, IEEE 7th International
Conference for Convergence in Technology, Mumbai, 2022, pp. 1-6, doi:
10.1109/I2CT5491.2022.9824896.
[11]George A Bray, William E Heisel, Ashkan Afshin, Michael D Jensen, William H Dietz, Michael
Long, Robert F Kushner, Stephen R Daniels, Thomas A Wadden, Adam G Tsai, Frank B Hu, John M

Page | 124 DOI:10.36893.DRSR.2023.V13I04.118-125 Copyright @ 2023 Authors

Dogo Rangsang Research Journal UGC Care Group I Journal
ISSN : 2347-7180 Vol-13, Issue-4, April 2023
Jackicic, Donna H Ryan, Bruce M Wolfe, Thomas H Inge, “The Science of Obesity Management: An
Endocrine Society Scientific Statement”, Endocrine Reviews, 2018, vol: 39, doi:
https://doi.org/10.1210/er.2017-00253.
[12]https://gist.githubusercontent.com/sukanyabag/fec853cee5c77ec362820e2131777911/raw/8438f
14bb7f48ba2f7dc8a7057e5bf23259a690a/LSTM.py
[13]https://www.nbshare.io/notebook/249468051/How-To-Code-RNN-and-LSTM-Neural-
Networks-in-Python/
[14] https://towardsdatascience.com/what-does-rmse-really-mean-806b65f2e48e
[15]https://medium.com/hackernoon/understanding-architecture-of-lstm-cell-from-scratch-with-
code-8da40f0b71f4**
[16]https://www.alpha-quantum.com/blog/long-short-term-memory-lstm-with-python/long-short-
term-memory-lstm-with-python/
[17]https://bond-kirill-alexandrovich.medium.com/lstm-for-real-time-recommendation-systems-
f5191d564be5
[18] https://towardsdatascience.com/building-a-random-forest-classifier-c73a4cae6781
[19]https://data.nal.usda.gov/dataset/usda-national-nutrient-database-standard-reference-legacy-
release

Page | 125 DOI:10.36893.DRSR.2023.V13I04.118-125 Copyright @ 2023 Authors