Smart Indian-Inspired Recipe Finder: Healthier

Culinary Choices
A V S S Sampath, Saurabh Kiran Mahuli, G Meher Pranav, V Kanapathi
Department of Computer Science and Engineering
Amrita School Of Computing, Bengaluru
Amrita Vishwa Vidyapeetham, India
[email protected],
[email protected],
[email protected],
[email protected],

Abstract—The vital element of survival in life is food. People Utilizes Positive Point-wise Mutual Information (PPMI) to
around the world love to cook, and knowing what to cook can record food preferences and evaluate the relation between
sometimes become trickier. The project we have built helps them food items in the dataset. Uses Machine Learning to provide
find recipes based on what they have and what they like to
eat. This search happens through an exclusive dataset primarily personalized nutrient recommendations for individuals with
focusing on cuisines from all over the world with an Indian touch. Type 2 Diabetes.[1]
Using sophisticated algorithms and profound data structures, The paper gives us a Detailed breakdown of Food Recom-
which are hashmaps and linked lists written in the language mender Systems (FRS) and future potential. FRS is cate-
of Java, an effective recipe is recommended to the user, keeping gorized by recommendations, user profiles, and algorithms.
in mind all the food criteria mentioned by the user. At the end,
a set of best-suited recipes is recommended to the users, with FRS utilizes machine learning, fuzzy logic, and deep learning.
the initial recipes being those that best meet the user’s criteria, Various algorithms like KNN, NLP, SOM, and PCA are used
gradually tapering off. This project aims for people to find it in FRS. It focuses on user behavior, preferences, and health
useful in finding food that meets their requirements and ensures in food recommendations. Concludes by explaining the nature
optimal health and wellness. of the recommendation system and its potential to transform
Index Terms—recipe, cuisine, recommendation system, user
criteria, personalized recommendations how individuals manage their diets in fast-paced lifestyles.[2]
This paper with the help of machine learning techniques,
I. I NTRODUCTION offers personalized food recommendations to users. This is
done by combining a content-based and collaborative filtering
Our food recommender is a one of a kind model that uses mechanism with the use of the K-nearest neighbour algorithm.
various different data structures to apply store and retrieve User inputs such as weather conditions, dietary restrictions,
recipes directly. We value the choice of the user to arrange and time of day are used to provide highly personalized food
recommendations in a certain priority and allow a more user suggestions.
friendly experience.We also look after the available inventory A combination of supervised and unsupervised learning meth-
to make sure that the recommended recipes can be followed. ods helps recommend diabetic-friendly foods, thereby showing
As you can see we have used a similar style as many other the importance of specific dietary needs.[3] This system rec-
papers where we take inspiration from an already available ommends Indian food recipes by examining their ingredients.
dataset to optimise out recipes. The literature survey also Information is gathered from different websites where recipes
shows the various machine learning algorithms that are ap- are published.
plied in learning models. Few of them have also considered The recipe text is understood by using a bag-of-words model,
the real time recipe reviews to provide better and realistic a common technique used in natural language processing. This
results. Collaborative filtering is also utilized to generate recipe system discusses the workings of the recommendation system
recommendations effectively. and also addresses the challenges encountered and proposes
ideas for future improvements.
II. L ITERATURE S URVEY
Tasks such as data cleaning, web scraping to gather data,
The paper introduces a unique approach to food recommen- and data processing are specifically modified for Indian cook-
dation using Multi-Objective Optimization through evolution- ing.[4]
ary computing, aiming to balance user preference, diversity This paper utilizes a Support Vector Machine (SVM) for
score, and nutrition score quantitatively. Recommends optimal suggesting cuisines based on the reviews of the restaurants.
food item combinations based on user preferences, dietary Many factors, such as customer preferences, sales trends,
restrictions, and daily nutritional requirements, emphasizing and potential product ideas, help analyse a particular review.
meal components like appetizers, main courses, and desserts. Meals are recommended using real-world data with a high
accuracy rate of 91 percent. Additionally, this paper discusses sources and analysis evaluation has also been conducted
the significance of recommendation systems across various which shows diversity in user preferences and accuracy of
industries. Training of the SVM classifier is done by utilizing a recommendations. User preferences influence the number of
dataset comprising 10,000 Yelp reviews, thereby showcasing a recipe recommendations received.[8]Personalised food recom-
practical application of machine learning algorithms in dietary menders are a necessity in todays day and age given the poor
recommendation systems.[5] eating habits of people.Personal model will remember crucial
This paper deals with a food recommendation system using data and recommend food based on inputs.Using multiple
K-nearest neighbor’s methods. The researchers have chosen models will help in better learning of the user and lead to
the Python programming language for implementation. They a more apt recommendation.[9] Riya Roy has successfully
have taken parameters like food id,cuisine type,diet type for a created Cosmo and Calypso, an AI powered robot and a rule
filtration recommendation system. The researchers have also based programming language that helped her in creating a
used collaborative filtering systems to rate attributes like food food recommendation system.The future goal is to combine
id etc. various API’s and allow entire food making process to take
The evaluation of the machine learning algorithm is done place through a single mobile application[10] This paper uses
using RMSE(Root Mean Square Error) which measures the a unique model called fuzzy inference system to suggest food
average squared difference between the predicted values and preferences based on socio-economic barriers. The system
the actual values. The second evaluation method which they utilises membership functions to determine levels of food
have chosen is the MAE(Mean Absolute Error) where this preference and tailors recommendations to match specific user
measures the average absolute difference.They have used 2 preferences and dietary needs of users.There are various levels
modes of filtration. The first being the Content-based filtering of food preferences that help sorting the given choices in an
which recommends food items based on textual information order.[11]
and the second being the collaborative filtering which suggests
food based on the user preferences and ratings. III. D ESIGN AND I MPLEMENTATION
These both systems collectively fulfill the result set.These
researchers have incorporated string matching to enhance the
recommendation process. Future work can include expanding
system capabilities with additional criteria.[6]
This paper focuses on healthy lifestyle,diet and customized
nutrition recommendations. This also addresses health issues
like obesity, diabetes and fast-food consumption. Recommends
nutrition-based smart recipes for healthy living. This paper
extensively utilizes Machine learning algorithms like K-means
algorithm, self-organizing map, genetic algorithm. Researchers
of this paper have used Long short-term memory networks to
predict heating curves, they have also used Ontology-based
web service to generate personalized healthy food charts. This
also generates diet plans based on the user’s health conditions
and recipe information. Researchers have also highlighted the
issue of diabetes by a proactive diabetes self-care recommen-
dation system.
This model focuses on accuracy in food recommendation using
Graph Convolutional Network. Also takes care about nutrient
balance while recommending a diet plan to the users.[7] This
paper implements a recipe recommendation system based on
ingredient availability with the user. Focuses on providing
recipe recommendations using content-based and collabora-
tive filtering.Utilizes content-based filtering to generate recipe
recommendations effectively.
The data gathered from various websites to provide popu-
lar Indonesian food recipes is evident from this paper. The
Researchers of this paper have used content-based filtering
for recipe similarity identification and recommendations, also
used matching methods like exact matching and use of Jac-
card similarity metric for comparing user’s ingredients with
recipe ingredients is also found. Researchers have used the Fig. 1. System Architecture of the Project
web scraping techniques to gather information from different Fig.1. explains the entire flow of the study.
 In the java code used in the project, the RecipeSearch class the results are displayed in the result section, which displays
reads and processes the recipe data from the CSV file. The the recipe name, cuisine, diet, course, matched ingredients, all
CSV file contains a dataset of recipes including details such ingredients, instructions, and relevance scores.
as recipe names,cuisines,ingredients,courses,diets and cooking
instructions. The CSV file is read using the OpenCSV java IV. R ESULT
library,which makes it easier to handle and parse data.
The dataset called ”6000 Indian Food Recipes Dataset” by We have observed that if we store the data in a linked list
Kanishk307, hosted on Kaggle, has a huge collection of Indian versus if we stored it in an array we observe that the retrieval
recipes, with over 6,000 of them to explore. The dataset has of data is much quicker in that of an array then that of a
recipes with the ingredients, cuisine, course type, and diet type linked list. We can say that this is mostly because of the fact
mentioned in them, with a diverse variety of recipes ranging that we are working majorly with retrieval and not a lot of
from the north to the south of the nation. inserting and deletion which improves the overall speed of
The recipes are then filtered based on the user-input such as the operation.
available ingredients,course of the recipe and the diet. The
kind of food that is eaten at a specific mealtime is specified V. F UTURE W ORK
by the course.Dinner, the main course, lunch, appetizers, side The future scope of this project could be to add more
dishes, etc. are a few examples. The term diet refers to dietary recipes from different cuisines and diets. An addition to
preferences, which are more similar to particular, important including the nutritional Information could help the user track
rules that must be followed when preparing food. These have their protein, carbohydrate, and calorie intake. A user rating
many different special instructions, like as vegan, high-protein, system would allow the user to rate the recommended recipes,
vegetarian, diabetic-friendly. and future suggestions would kind of depend on the ratings
A hashmap is used to store each recipe’s data and a linked given. Taking this project to a further level could be real-life
list to maintain the list of recipes that match the user’s search deployment in a mobile application, which could provide a
criteria. more easy and hassle-free experience. The futuristic approach
The hashmap is preferred for this tool because it helps with could be to add voice commands to make it more convenient,
easy traversal based on keys. Each recipe can be stored as especially for people with disabilities.
a key-value pair, where the key contains the column name
and the value has the corresponding attribute of the column.
Hashmaps provide a constant time complexity of O(1) for VI. C ONCLUSION
insertion and search operations. This is a great feature of
this data structure, which helps in handling large datasets and The recipe recommendation system developed in this project
provides quick access to the data in the large dataset. provides an optimal and user-friendly solution for finding
The linked list is another best choice to maintain the list recipes based on the user’s available resources, course type,
of recipes that match the user’s criteria. This data structure and dietary preferences. With the help of Java programming
provides efficient insertion and deletion operations with a time language, data structures and algorithms, the system can search
complexity of O(1) if the head and tail pointers are maintained. through a large dataset of recipes and present the best recom-
If a recipe matches the user criteria, it can be quickly added mendations to the user. The combination of a HashMap and
to the list without resizing or shifting elements, as the linked a linked list data structure ensures efficient storage, retrieval,
list can grow and shrink dynamically. and ranking of recipes based on the user input. The HashMap
After the right choice of data structures is made, the relevance allows access to recipe details, while the linked list allows us
score of the recipe is calculated based on the number of to add and remove recipes that match the search criteria. The
matched ingredients compared to the total available ingredients relevance scoring algorithm helps to rank the recommended
mentioned by the user. This assists in ranking the recipes in recipes based on their ingredient match, ensuring that the most
accordance to the relevancy. Only those recipes are displayed relevant options are presented first.
whose relevance score is greater than or equal to the relevance The use of the RecipeSearchGUI class provides a user-friendly
threshold. This relevance threshold is a constant value which is user interface, allowing users to input their preferences easily
set to 0.4 which provides a perfect balance in recommending and view the recommended recipes in a structured and visually
the recipes At the end the matched recipes are sorted according appealing manner. Java Swing is used for the GUI develop-
to their relevance scores and are displayed in a descending ment.
order to meet the criteria of highest relevancy first. Overall, this recipe recommendation system demonstrates the
The class RecipeSearchGUI provides a user-friendly graphic implementation of data structures, algorithms and program-
user interface (GUI) implemented with the help of Java Swing, ming techniques to solve a problem in a regular household and
and this tool contains sections such as input fields, which are professional restaurants. By considering the user’s preferences
text fields for users to enter the ingredients, course type, and and available resources, the system tells the individual to
dietary choice. When the search button is clicked, the backend explore new recipes while ensuring optimal usage of their
logic is initialized to search through the CSV. After searching, ingredients while aligning with their dietary needs.
 R EFERENCES
[1] Y. Perera, R. Karannagoda, D. Weiman and S. Fernando, ”Mul-
tiple Objective Optimization Based Dietary Recommender System,”
2023 8th International Conference on Information Technology Re-
search (ICITR), Colombo, Sri Lanka, 2023, pp. 1-6, doi: 10.1109/IC-
ITR61062.2023.10382927.
[2] M. Ashraf, S. S. Sohail and B. M. Chaudhry, ”Issues, Challenges,
and Opportunities in Food Recommender Systems,” 2022 Interna-
tional Conference on Data Analytics for Business and Industry
(ICDABI), Sakhir, Bahrain, 2022, pp. 570-574, doi: 10.1109/ICD-
ABI56818.2022.10041496.
[3] K. A, V. S, D. R and M. Balamurugan, ”Smart Dine-in: A Personalized
Food Recommendation System,” 2023 Intelligent Computing and Con-
trol for Engineering and Business Systems (ICCEBS), Chennai, India,
2023, pp. 1-6, doi: 10.1109/ICCEBS58601.2023.10448742.
[4] N. Nilesh, M. Kumari, P. Hazarika and V. Raman, ”Recommendation
of Indian Cuisine Recipes Based on Ingredients,” 2019 IEEE 35th
International Conference on Data Engineering Workshops (ICDEW),
Macao, China, 2019, pp. 96-99, doi: 10.1109/ICDEW.2019.00-28.
[5] A. Pavate, A. Chaudhary, P. Nerurkar, P. Mishra and M. Shah, ”Cuisine
Recommendation, Classification and Review Analysis using Supervised
Learning,” 2020 International Conference on Convergence to Digital
World - Quo Vadis (ICCDW), Mumbai, India, 2020, pp. 1-6, doi:
10.1109/ICCDW45521.2020.9318646.
[6] R. Singh and P. Dwivedi, ”Food Recommendation Systems Based On
Content-based and Collaborative Filtering Techniques,” 2023 14th In-
ternational Conference on Computing Communication and Networking
Technologies (ICCCNT), Delhi, India, 2023, pp. 1-5, doi: 10.1109/IC-
CCNT56998.2023.10307080.
[7] L. A. Shubhashree, S. Chaudhari and R. Aparna, ”A Nutrition-Based
Smart Recipe Recommender For Healthy Living,” 2022 IEEE 3rd Global
Conference for Advancement in Technology (GCAT), Bangalore, India,
2022, pp. 1-6, doi: 10.1109/GCAT55367.2022.9971930.
[8] Y. W. Syaifudin, T. Fatmawati, N. A. Ardyningrum, I. Siradjuddin,
D. Puspitasari and M. Astiningrum, ”An Implementation of Recipe
Recommendation System Based on Ingredients Availability Using
Content-based and Collaborative Filtering,” 2023 International Confer-
ence on Converging Technology in Electrical and Information Engineer-
ing (ICCTEIE), Bandar Lampung, Indonesia, 2023, pp. 96-101, doi:
10.1109/ICCTEIE60099.2023.10366631.
[9] W. Min, S. Jiang and R. Jain, ”Food Recommendation: Frame-
work, Existing Solutions, and Challenges,” in IEEE Transactions
on Multimedia, vol. 22, no. 10, pp. 2659-2671, Oct. 2020, doi:
10.1109/TMM.2019.2958761.
[10] R. J. Roy, ”A Futuristic Kitchen Assistant – Powered by Artifi-
cial Intelligence and Robotics,” 2020 IEEE Integrated STEM Edu-
cation Conference (ISEC), Princeton, NJ, USA, 2020, pp. 1-5, doi:
10.1109/ISEC49744.2020.9397856.
[11] M. Panwar, A. Sharma, O. P. Mahela and B. Khan, ”Fuzzy Inference
System Based Intelligent Food Recommender System,” 2023 3rd Asian
Conference on Innovation in Technology (ASIANCON), Ravet IN,
India, 2023, pp. 1-6, doi: 10.1109/ASIANCON58793.2023.10270048.