INTERNATIONAL JOURNAL OF RESEARCH AND INNOVATION IN SOCIAL SCIENCE (IJRISS)

ISSN No. 2454-6186 | DOI: 10.47772/IJRISS | Volume IX Issue X October 2025

AI-Powered Personalized Allergen Detection and Recipe

Modification Tool for Safer Meal Preparation

Preevitha Nithiyananthan^1,Kurk Wei Yi², Emaliana Kasmuri³, Fadhlan Faizal bin Nuruddin⁴

^1,2,3Fakulti Teknologi Maklumat dan Komunikasi, Universiti Teknikal Malaysia Melaka, Durian

Tunggal, Melaka, 76100, Malaysia

⁴GB Multimedia Sdn Bhd, FR-03M-02, Tamarind Square, Persiaran Multimedia, Cyber10, 63000

Cyberjaya, Selangor, Malaysia

DOI: https://dx.doi.org/10.47772/IJRISS.2025.910000555

Received: 28 October 2025; Accepted: 04 November 2025; Published: 18 November 2025

ABSTRACT

The AI-Powered Personalized Allergen Detection and Recipe Modification Tool is a mobile application

developed to assist individuals with food allergies helping them identifying harmful ingredients in recipes and

offer safer alternatives tailored to their personal allergen profiles. Food allergy is a serious health concern that

can lead to life-threatening reactions if not managed properly, especially when consuming meals with unfamiliar

ingredients. This tool uses artificial intelligence (AI) and natural language processing (NLP) to analyse the

textual format recipes specified by users and intelligently detect the presence of any allergen substances. Once

allergens are identified, the tool accesses a built-in database integrated with machine learning/large language

model to suggest appropriate and non-allergenic substitutes, allowing the users to prepare and modified a safer

recipe. The tool includes features such as creating, editing, sharing, favouriting, and deleting recipes, along with

user account settings that support allergen input and password management. With a focus on 90% accuracy rate

in allergen detection and 44% substituion relecance score, the app ensures both speed and reliability. The tool

was developed specifically for Android platform and intended for online use has provide an intuitive and user-

friendly interface designed for convenience and efficiency. The tool helps users to make informed decisions

about the ingredients and substitutions and early user feedback indicates improved confidence in meal

preparation.

Keywords: Food Allergy, Allergen Detection, Recipe Modification, Artificial Intelligence (AI), Mobile Health

Application

INTRODUCTION

Food allergies are a growing global public-health concern, affecting both children and adults. They occur when

the immune system misidentifies certain food proteins as harmful, causing symptoms that can range from mild

irritation to life-threatening anaphylaxis [1]. The U.S. Food and Drug Administration estimates that millions of

people worldwide experience food-related allergic reactions annually, and that strict avoidance of the offending

foods remains the only reliable preventive measure [1]. In the United States alone, approximately 6 percent of

adults and 8 percent of children live with at least one diagnosed food allergy. The “Big 9” allergens: milk, eggs,

peanuts, tree nuts, wheat, soy, fish, shellfish, and sesame are responsible for most severe reactions [2].

Despite labelling regulations and increased public awareness, individuals with food allergies continue to face

difficulties when preparing or consuming meals. Hidden allergens in complex or processed foods are especially

problematic, and manual ingredient checking is both time-consuming and error prone [3]. Traditional recipe or

nutrition applications typically provide static ingredient databases and do not support real-time allergen detection

or personalized substitution recommendations. These shortcomings restrict dietary variety and elevate the risk

of accidental exposure.

Recent studies have shown artificial intelligence (AI) and natural language processing (NLP) are able to analyze

Page 6788

www.rsisinternational.org

INTERNATIONAL JOURNAL OF RESEARCH AND INNOVATION IN SOCIAL SCIENCE (IJRISS)

ISSN No. 2454-6186 | DOI: 10.47772/IJRISS | Volume IX Issue X October 2025

ingredient list, detecting allergen presence and suggest safe substitution automatically for intelligent dietary

management system [4]. Studies have demonstrated AI’s potential to improve allergen detection accuracy,

enhance dietary planning, and support patient decision-making in real-time environments [5]. However, few

mobile applications integrate these intelligent capabilities with user-centred interfaces that allow recipe creation,

allergen analysis, and ingredient modification within a single tool.

This study develop the tool using React Native and MySQL backend that uses AI -based API to identify allergens

and suggest safer substations. The tool enables users to input and edit recipes, analyses ingredients using NLP

for allergen detection, and proposes appropriate, safe substitutions. The integration between AI technology and

mobile accessibility shall empower individuals with food allergies in preparing safer meals, promotes a healthier

and more inclusive dietary practices [4], [6].

Background

Food allergies are increasingly recognized as a major global health issue with a rising prevalence across all age

groups. Recent studies highlight that the incidence of both adult and infant food allergies has continued to grow

worldwide, primarily due to changes in diet, food processing, and environmental factors [7]. Despite ongoing

efforts to improve labelling and awareness, accidental allergen exposure remains common in daily food

preparation and consumption. Conventional recipe databases and nutrition-tracking applications focus largely

on nutrient values rather than allergen safety, leaving users to manually identify allergens, a process prone to

error and inefficiency.

Advancements in artificial intelligence (AI) and machine learning (ML) have provided promising tools for

improving food-safety monitoring and allergen detection. AI algorithms can analyse ingredient-level data,

identify hidden allergenic proteins, and predict potential reactions using pattern-recognition models. Recent

work by Yang et al. [8] demonstrated a novel AI-driven method using near-infrared spectroscopy (NIRs) for

early detection of non-specific lipid transfer protein (nsLTP) allergens, enabling fast and non-destructive

screening. Similarly, Li et al. [9] developed a portable fluorescence biosensing system enhanced with AI, capable

of detecting multiple allergens simultaneously, marking a significant leap toward real-time, point-of-care

allergen identification.

Beyond detection, the integration of AI into personalized nutrition has fostered the rise of precision nutrition an

approach that tailors dietary recommendations to an individual’s genetic, physiological, and lifestyle factors.

Deep learning models that combine microbiome and diet data have demonstrated effectiveness in predicting

optimal nutrition strategies and allergy risk profiles [10]. These developments lay a strong foundation for

intelligent applications that not only identify allergens but also assist users in modifying recipes according to

their unique health needs.

The adoption of mobile health (mHealth) technology further enhances accessibility and real-time feedback in

dietary management. Smartphones serve as effective platforms for hosting AI-powered food-safety applications,

allowing users to receive instant allergen detection and substitution suggestions. However, most existing mobile

apps are limited to static allergen lists and lack adaptive learning capabilities [11]. Thus, integrating AI and NLP

in mobile health technologies allowed our tool to provide personalized allergen detection that improves the user’s

confidence in their food choices [8], [9].

Related Work

Recent developments in artificial intelligence (AI) and machine learning (ML) have revolutionized food safety,

quality assurance, and health-related analytics. The adoption of AI-driven systems enables automated data

processing and real-time decision-making, particularly in detecting contaminants, allergens, and nutritional

irregularities. Revelou et al. [12] provided a comprehensive review of AI and ML applications in food safety

and quality control, emphasizing their success in industrial environments for contamination detection and

inspection automation. However, their work focused primarily on large-scale production systems and lacked

attention to personalized allergen management.

Page 6789

www.rsisinternational.org

INTERNATIONAL JOURNAL OF RESEARCH AND INNOVATION IN SOCIAL SCIENCE (IJRISS)

ISSN No. 2454-6186 | DOI: 10.47772/IJRISS | Volume IX Issue X October 2025

A growing body of research now explores AI’s role in allergen detection and classification. Sarlakifar et al. [13]

introduced AllerTrans, a deep learning (DL) framework designed to predict protein allergenicity using sequence-

level features. The model demonstrated improved precision and recall rates for allergen prediction compared

with conventional classifiers. Similarly, Liu et al. [14] proposed a computational model integrating multiple

feature types, such as sequence patterns, physicochemical properties, and evolutionary profiles to improve

allergenic protein prediction accuracy. Despite these advances, both studies were confined to laboratory-based

datasets and did not address real-world dietary applications or mobile implementation for consumers.

Kim et al. [15] extended the application of AI to food safety and contamination control by incorporating

computer vision and ML algorithms into automated quality inspection systems. Their findings reinforced AI’s

reliability in ensuring food integrity throughout the supply chain. Yet, this research remains limited to industrial

settings and lacks a personalized dietary component.

Meanwhile, Chen et al. [10] explored the concept of precision nutrition through deep learning techniques applied

to microbiome and dietary data. Their study demonstrated that AI can deliver tailored nutrition recommendations

to improve health outcomes, but the framework did not consider allergen avoidance or recipe adaptation for

allergic individuals.

Collectively, these works affirm the growing importance of AI in food safety and health personalization.

However, most of the reviewed studies focus either on industrial quality control or theoretical allergen modelling

rather than user-centric applications. There remains a substantial gap in research concerning the integration of

AI-based allergen detection, NLP-driven recipe analysis, and mobile accessibility within a single, interactive

platform.

Table1 Summary of Related Works

Ref

Method / Approach

Key Findings

Identified Research Gap

Review of ML and AI Provided an overview of how Focused on industrial food

methods applied to food AI improves food-quality inspection; no personalization or

[12]

inspection

contamination detection

and assurance

monitoring

contexts

and

in

safety

industrial

allergen-specific application

Deep

learning

model Developed a DL framework Limited to laboratory datasets;

real-world, user-level

integration or mobile deployment

[13]

[14]

(AllerTrans) for protein with higher accuracy for lacks

allergen classification

predicting allergenic proteins

Multi-feature

fusion Achieved improved allergen Computational focus; no recipe-

(sequence,

physicochemical,

evolutionary features)

prediction through

and

feature data fusion

multi- level or dietary personalization

component

AI-driven

chemical

inspection

visual

food-quality automating

control and quality assessment

and Demonstrated AI’s role in Industrial and product-level

[15]

[10]

contamination orientation; does not support

personalized

management

food-safety

nutrition

Deep

learning

on Established AI frameworks for Focused

on

microbiome and dietary individualized

diet personalization; no allergen

health detection or recipe modification

functionality

optimization

improvement

and

data

Page 6790

www.rsisinternational.org

INTERNATIONAL JOURNAL OF RESEARCH AND INNOVATION IN SOCIAL SCIENCE (IJRISS)

ISSN No. 2454-6186 | DOI: 10.47772/IJRISS | Volume IX Issue X October 2025

METHODOLOGY

The Allergen Detection and Recipe Modification Tool was developed using a systematic agile software

development process. The process consist of four major phases which are analysis, design, implementation and

testing. The implementation and testing were executed iteratively ensuring the functional and non-functional

requirements effectively translated into mobile solution that is designed to identify allergens, recommend

substitution and enhancing user safety in meal preparation.

Analysis Phase

The analysis phase is aimed to understand the overall behaviour of the tool. The Allergen Detection and Recipe

Modification Tool requirements are modelled using the use case diagram presented in Fig 1. The main actor of

application is the User, who interacts with the system to perform several key actions. These include registering an

account, customizing recipes, tracking recipe trials, and sharing customized recipes with others. These core

functions represent the primary ways users engage with the application to manage their food allergies and

personalize their cooking experience.

Fig1. Use case diagram proposed application

A central use case in the diagram is Analyze Recipe, which captures the system’s ability to examine the

ingredients of a recipe provided by the user. This analysis is essential for identifying potential allergens based on

the user’s personalized allergy profile. Extending from this use case is Suggest Substitute Ingredient, which is

triggered when allergens are detected. This optional feature provides safe and suitable ingredient alternatives,

allowing users to modify their recipes effectively. The diagram emphasizes the tool’s user-centric design, focusing

on practical features that support safe and informed meal preparation while promoting ease of use and

personalization.

Design Phase

The design phase focused on transforming the tool requirement that was previously defined in the analysis stage

into an architectural framework that support efficient data processing, scalability, and user accessibility. The

architectural framework consist of several modules as illustrated Fig 2.

Fig 2.The Allergen Detection and Recipe Modification Tool architectural workflow

Page 6791

www.rsisinternational.org

INTERNATIONAL JOURNAL OF RESEARCH AND INNOVATION IN SOCIAL SCIENCE (IJRISS)

ISSN No. 2454-6186 | DOI: 10.47772/IJRISS | Volume IX Issue X October 2025

The framework begins with Recipe Creation module. This module interfacing with the user to accept the recipe.

The user will send the recipe for allergen analysis through an API Gateway. This gateway acts as a centralized

entry point, efficiently directing traffic to the appropriate backend services.

The Allergen Detection module responsible to analyse the user recipe. The module is the most critical

component in the architecture where ingredients are identified, extracted and analysed using an AI/ML model.

The AI/ML model is an encapsulated large language model (LLM) library consist of intelligent algorithms

trained various text documents. The AI/ML model determined the allergen presence in the recipe based on the

user dietary and allergen profile. The identified ingredients are compiled to be processed in the next module.

The Substitution Recommendation module uses the AI/ML model to provide safe ingredient options based on

the identified user allergenic ingredients for the user. The options were gathered and presented to the user. The

tool is designed to allow user to select the recommended ingredients instead of deciding for the user.

The Recipe Customization module allow the user to substitute the allergenic ingredients with the recommended

ingredients. The module manage the customized recipe and saves it the database for the user future’s reference

and use.

The Recipe Trial Tracking module records the number of user attempt and feedback. The feedback serves as a

continuous allergy monitoring loop for the user enhancing meal preparation safety.

Development Phase

The implementation phase marks the transition from design to a functional tool for the Personalized Allergen

Detection and Recipe Modification Tool. This phase focused on translating the architecture framework, data

design, and user interface design into an operational mobile application. The development began with setting up

the software environment using React Native for cross-platform front-end development and MySQL as the

backend database. Tools such as Android Studio, Visual Studio Code, and MySQL Workbench were configured

to streamline development, testing, and debugging processes. The API layer which is hosted at

http://67.205.160.200:5000, acted as the bridge between the mobile app front-end and backend and managing

communication for all modules described in Fig 2.

Each module was implemented in iteration, adhering to the Agile principle. The iteration modularised the

development artefacts that results in ease of testing, debugging and maintenance. The Account Registration and

Authentication modules were first to be developed. These modules allowed users to securely create and access

accounts using encrypted credentials. Once these modules were established, the Recipe Creation modules were

built to handle recipe creation, editing, deletion, and sharing. Next, the Allergen Detection module was

developed and integrated into the development. Then followed by Substitution Recommendation module. In

these two module, access to AI/ML module was embedded, realizing the safer approach in meal preparation.

The Recipe Customization module were added after that, realizing the personalization aspect of the tool. Finally,

the Recipe Trial Tracking module were integrated into the tool to rate recipes, indicate preparation difficulty,

and report whether substitutions were effective or safe, enhancing personalization and tool learning.

Version control was maintained throughout development using a systematic commit and review process,

ensuring stable integration of new features and allowing rollback when necessary. The frontend application

communicated through an API gateway to the backend services, including user management, recipe

management, and allergen detection services. These components interacted with the user and recipe databases

to store and retrieve data efficiently.

Testing Phase

The testing phase aimed to validate the functionality, performance, and usability of the Allergen Detection and

Recipe Modification Tool, ensuring it met user requirements and project specifications. A structured testing

process was followed, combining functional, integration, and user acceptance testing (UAT). This systematic

approach allowed for early defect identification, iterative correction, and the delivery of a stable and user-friendly

application.

Page 6792

www.rsisinternational.org

INTERNATIONAL JOURNAL OF RESEARCH AND INNOVATION IN SOCIAL SCIENCE (IJRISS)

ISSN No. 2454-6186 | DOI: 10.47772/IJRISS | Volume IX Issue X October 2025

The testing were focusing on two critical modules which are the Allergen Detection and Substitution

Recommendation. The test cases for these modules are describe in Table II. The testing was conducted in a

controlled environment comprising the developer’s workstation and user mobile devices. The tool APK was

installed on Android devices running version 8.0 and above, with evaluations conducted under both Wi-Fi and

4G/5G mobile networks to simulate real-world conditions. Tools such as Android Studio, MySQL Workbench,

and Microsoft Survey were employed for functional validation, data management, and feedback collection.

Manual test documentation and bug-tracking logs were used to record results and ensure traceability across all

testing cycles.

Table 2. Test Cases

Test Case

Scenario

Input Condition

Expected Output

Recipe:

Peanut

Butter

Peanut is detected and highlighted.

TC01

Basic Allergen Detection

Sandwich

Recipe: Creamy Pasta with Peanut and milk are detected

TC02

Multi-Allergen Detection

Peanut Sauce (contains milk

& peanuts)

highlighted.

Hidden

Compound Ingredient

Allergen

in

Recipe: Soy Sauce Chicken

TC03

TC04

TC05

TC06

TC07

Soy is detected.

Recipe: Pad Thai (contains

fish sauce, peanuts)

Nested Ingredient Parsing

Fish sauce and peanut are detected.

Substitution Recommendation

– Simple

Sunflower seeds or pumpkin seeds

are recommended.

Recipe with peanuts

Substitution Recommendation

– Multi-Allergen

Oat milk and, sunflower seeds are

recommended.

Recipe with milk & peanuts

Recipe: Thai Peanut Sauce

Almond butter or tahini is

recommended to maintain texture

Culinary Intent Preservation

TC08

TC09

Regional Ingredient Parsing

Complex Recipe Parsing

Recipe: Roti Canai with Ghee Ghee is detected from milk allergen

Recipe: Vegetarian Sushi

Soy and egg allergens detected

(contains

soy

sauce,

mayonnaise)

Adaptive Substitution for Recipe:

Complex Dish

Seafood

(contains shellfish, fish stock)

Paella Plant-based seafood alternatives

and vegetable broth are

recommended.

TC10

The test results demonstrated that all test cases described in Table II are compiled and recorded in Table III. The

Allergen Detection module performed at 90% accuracy for correctly detecting allergen in the recipe. The module

perform at 92% success rate for correctly handling multiple allergen in a recipe. The high success rates shows

the module robustness handling single and multiple allergen detection in a recipe. Apart from explicit allergens

presence, the module was tested to complex ingredient. The module performed at 85% accuracy on complex

ingredient, showing a competitive advantage to the module.

The relevance score metrics is used to measure the contextual appropriateness and correction for the Substitution

Recommendation module. The module achieved 47% score, indicating the tool struggle to maintain the original

recipe intent with the substituted ingredient.

Page 6793

www.rsisinternational.org

INTERNATIONAL JOURNAL OF RESEARCH AND INNOVATION IN SOCIAL SCIENCE (IJRISS)

ISSN No. 2454-6186 | DOI: 10.47772/IJRISS | Volume IX Issue X October 2025

Table3. Testing Results

Metrics

Description

Result

90%

92%

85%

44%

Accuracy

Correct allergen detection

Multi allergen success rate

Complex Ingredient Parsing Accuracy

Substitution Relevance Score

Correct handling of multiple allergens

Correct detection in compound ingredients

Culinary intent preserved in suggestions

Other modules which are Recipe Creation, Recipe Customization and Recipe Trial Tracking are tested during

User Acceptance Test (UAT). The UAT results shows the app is perceived as highly effective in accidental

allergen exposure prevention with 70 Net Promoter Score (NPS), as illustrated in Fig 3. The UAT involves 30

respondents, where 21 promoters were recorded, indicating strong satisfaction on the tool and the likelihood to

recommend or influence other to user the tool is high. While 9 respondents were recorded as passives, indicating

moderately satisfy with the tool even though the likelihood to recommend or influence other to use the tool is

low. Importantly, there were no detractors, which reflects a positive overall user experience. The high NPS

score demonstrating a strong trust in the app capability analysing, identifying allergen presence and substitution

recommendation in different degree of recipe complexity. Furthermore, the absence of major defects during

UAT, confirming the tool’s stability and reliability for deployment.

Fig3. Net promoter score of proposed application

In summary, the testing phase validated that the developed tool achieved its objectives of accurate allergen

detection, intuitive recipe modification, and high usability. The combined results from functional and user

acceptance testing confirm that the AI-Powered Personalized Allergen Detection and Recipe Modification Tool

is ready for real-world use, offering a reliable and effective solution for food allergy management and safe recipe

customization.

RESULT

The results presents the outcomes of the tool implementation, highlighting the actual user interfaces and

functionalities of the Allergen Detection and Recipe Modification Tool. The developed tool is a mobile app that

integrates user authentication, allergen detection, ingredient substitution, and feedback management within a

single responsive interface. The screenshots captured during testing and deployment to visually demonstrate the

proper functioning of each module according to its intended purpose.

User Authentication and Onboarding Modules

A new user will be guided through a secure onboarding process. The Sign-Up Screen in Fig 4 allows the user to

create an account by providing essential details including as full name, email address, and password. Upon a

success onboarding process, the registered user can access the tool using the Log In Screen as shown in Fig 5.

Page 6794

www.rsisinternational.org

INTERNATIONAL JOURNAL OF RESEARCH AND INNOVATION IN SOCIAL SCIENCE (IJRISS)

ISSN No. 2454-6186 | DOI: 10.47772/IJRISS | Volume IX Issue X October 2025

Fig. 1. Registration screen

Fig. 2. Login screen

The user may recover a forgotten password using the screen shown in Fig 6. Subsequently, a one-time password

will be emailed to the user to be provided in the screen in Fig 7. The OTP facilitate a secure password reset

workflow ensuring user data remains protected.

Fig. 4. OTP verification screen

Fig. 3. Forget password screen

Core Application Dashboard

Upon success authentication, users are directed to the Home Screen, shown in Fig 8. The Home screen used an

intuitive navigation design which featuring a search bar, a personalized welcome message, quick-access modules

to "Favourites," user-created recipes, and community feedback. A bottom navigation bar ensures seamless

movement between the application's primary sections.

Page 6795

www.rsisinternational.org

INTERNATIONAL JOURNAL OF RESEARCH AND INNOVATION IN SOCIAL SCIENCE (IJRISS)

ISSN No. 2454-6186 | DOI: 10.47772/IJRISS | Volume IX Issue X October 2025

Recipe Creation

The central functionality of the tool is the recipe creation. The tool provides two approaches to create a recipe.

The first approach is using a standard form as shown in Fig 9. In the form, the user is required to specify the

recipe title, description, supporting image and a list of ingredient.

Allergen Detection and Substitution Recommendation

The tool analyse the recipe and identify the allergen presence in the ingredient. Subsequently, the tool proactively

display the Substitute Ingredients screen shown in Fig 10 after it detects allergenic ingredient that best march

the user’s allergen profile. The screen suggest safe alternative ingredient to substitute with the allergenic

ingredients. The tool allow the user to customize the recipe using the listed substitute ingredient.

Fig. 5. Substitute ingredients

Fig. 6. Updated recipe screen

The tool display the customized recipe as shown in Fig 11. It is essential for the tool to confirmed with the user

on the customized recipe before automatically saved it into the recipe repository. The confirmation adds another

safety layer in food allergy management and meal preparation.

Recipe Management and Sharing

The tool offers comprehensive feature for managing personalized recipes. All saved recipes are accessible from

the Saved Recipe Screen shown in Fig 12, which includes a search function. Selecting a recipe will provides a

detailed view, clearly distinguishing between original and substituted ingredients, and offers options to favourite,

share, or delete. The Share Recipe Screen (Figure 9) allows users to easily distribute their safe, customized

recipes through other platforms like WhatsApp and email.

Fig. 7. Save and share recipes screen

Page 6796

www.rsisinternational.org

INTERNATIONAL JOURNAL OF RESEARCH AND INNOVATION IN SOCIAL SCIENCE (IJRISS)

ISSN No. 2454-6186 | DOI: 10.47772/IJRISS | Volume IX Issue X October 2025

User Feedback and Profile Customization

User feedback to track the effectiveness of the substituted ingredients used in the customized recipe is a critical

feature. The tool provide a mean of tracking the user attempt or trial on the customized recipe as shown in Fig

13.

The user maintain their account information using the screen in Fig 14. In some cases, the user might develop

a new food allergy or may no longer allergic to certain kind of food. The tool include the food allergy dynamics

that allows the user to update the allergenic food profile using the screen in Fig 15. The screen is accessible

from Fig 14. The screen in Fig 15 shall maintain the most up to date allergenic food information. Apart from

ensuring a safe meal preparation, the tool shall ensure the user essential nutrient according the user’s current

allergy condition.

Additional tool safety measure is provided which allowing the user to update the password from time to time

using the screen shown in Fig 16.

Fig. 8. Feedback integration screen

Fig. 10. Allergens update screen

Fig. 9. Account update screen

Fig. 11. Change password screen

Fig. 12. Favorite recipes screen

Favoriting a recipe is another safety measure in meal preparation and food allergy management. The favorite-

marked recipe is an allerged-checked and safe recipe that shall minimize accidental exposure to risky ingredient.

Page 6797

www.rsisinternational.org

INTERNATIONAL JOURNAL OF RESEARCH AND INNOVATION IN SOCIAL SCIENCE (IJRISS)

ISSN No. 2454-6186 | DOI: 10.47772/IJRISS | Volume IX Issue X October 2025

Besides that, a favorite-marked recipe provide quick access to the recipe leads to efficient time planning in

preparing the meal. Furthermore, it this feature encourage better user engagement with the tool.

CONCLUSION

The development of the Allergen Detection and Recipe Modification Tool successfully achieved its primary

goal of creating an intelligent mobile app solution that assists users in identifying potential allergens in food

ingredients and suggesting safe alternatives. The tool integrates mobile app technology, structured user allergen

profile, AI and NLP model to analyse the user recipes. The integration has consistently identified allergens and

recommend substitute ingredients with high reliability during testing. The completed tool aligns with the

objectives established during the analysis and design stages, proving to be both technically feasible and

functionally effective.

Throughout the project, each development phase—from analysis to testing was systematically executed. The

analysis phase defined clear user requirements, focusing on allergen safety and dietary personalization. The

design and implementation phases translated these requirements into a robust architectural workflow, ensuring

modularity and scalability. The testing phase validated all core functionalities through black-box and functional

testing, where all test cases achieved successful results. The results phase, supported by actual tool screenshots,

confirmed that each module: registration, login, allergen detection, substitution, feedback, and recipe creation

operated according to specification, demonstrating the tool’s readiness for real-world deployment.

The tool contributes to the broader field of health-aware food technology, emphasizing preventive allergen

management through mobile app solution. Its success lies not only in accurate allergen identification but also in

promoting safe recipe adaptation, helping users maintain their preferred diets while avoiding allergenic risks.

The application’s design ensures accessibility for users with varying technical literacy and provides a framework

that can be enhanced for larger databases or regional food variations.

In conclusion, the Allergen Detection and Recipe Modification Tool has proven to be a reliable, efficient, and

user-friendly platform that supports personalized allergen management. The combination of intelligent detection,

substitution logic, and real-time interaction demonstrates how modern technology can improve food safety and

quality of life for individuals with food allergies. This project establishes a solid foundation for future research

and development in mobile health applications that integrate AI-driven decision-support tools for dietary

management.

Future Works

For future development, we plan to improve the tool accuracy and privacy feature while exploring clinical

collaboration feature to increase its practical use. First, adding robust image-based food recognition feature

would greatly reduce the manual entry effort and improve usability. In the recent reviews and empirical work,

the deep learning models has demonstrated its capability to achieve high accuracy for food item recognition and

volume estimation [16]. Besides that, the models enables automatic ingredient extraction from photos for

subsequent allergen analysis [16], [17].

Second, integrating large language models (LLMs) or fine-tuned transformer models for ingredient

understanding and substitution can improve the quality and contextual relevance of replacement suggestions;

recent experiments demonstrate LLMs’ ability to propose nutritionally informed and phytochemical-aware

ingredient substitutions that preserve culinary intent while meeting dietary constraints [18]. The user privacy can

be enhanced in the future updates by integrating federated learning in the tool. This integration will enable the

tool to learn from user interaction without collecting personal data explicitly [19].

A feedback loop process integrated with reinforcement learning technique should be established in the future to

improve user engagement with tool. Besides that the data collected from the previous screen should be use to

analyse the user preference and dynamics of food allergy to improve a safer recommendation. Furthermore, the

access to variety of recipes in the database shall provide allergen-approved and safe options to the user.

Page 6798

www.rsisinternational.org

INTERNATIONAL JOURNAL OF RESEARCH AND INNOVATION IN SOCIAL SCIENCE (IJRISS)

ISSN No. 2454-6186 | DOI: 10.47772/IJRISS | Volume IX Issue X October 2025

Furthermore, expanding multilingual label and ingredient parsing is essential for global deployment: building or

adopting multilingual OCR and NLP pipelines (and curated multilingual ingredient datasets) will reduce false

negatives caused by non-English labels and localized ingredient names, improving detection across cuisines and

regions [16], [17].

Finally, coupling the application with user-centric personalization, such as adaptive recommendation engines

that learn from individual feedback, clinical severity, and (optionally) microbiome or wearable data can make

substitution suggestions and alerts more clinically relevant and acceptable to users, while careful user consent

flows and transparent explanations (explainable AI) will help maintain trust and safety. Together, these

enhancements would move the app from a rule-based detector to an intelligent, privacy-aware, and globally

usable decision-support tool for allergy-safe eating.

ACKNOWLEDGEMENT

The authors would like to express gratitude to Fakulti Teknologi Maklumat dan Komunikasi (FTMK), Universiti

Teknikal Malaysia Melaka (UTeM) for their invaluable support and resources provided throughout this research.

REFERENCES

1. U.S. Food and Drug Administration. (n.d.). Food allergies. Retrieved October 15, 2025, from

https://www.fda.gov/food/nutrition-food-labeling-and-critical-foods/food-allergies

2. U.S. Department of Agriculture, Food Safety and Inspection Service. (n.d.). Food allergies: The “Big

9.” Retrieved October 15, 2025, from https://www.fsis.usda.gov/food-safety/ safe-food-handling-and-

preparation/food-safety-basics/food-allergies-big-9

3. Food Allergy & Anaphylaxis Connection Team (FAACT). (n.d.). The basics for food allergies. Retrieved

October 15, 2025, from https://www.foodallergyawareness.org/food-allergy-and-anaphylaxis/food-

allergy-basics/food-allergy-basics/

4. Jiang, F., Jiang, Y., Zhi, H., Dong, Y., Li, H., Ma, S., Wang, Y., Dong, Q., Shen, H., & Wang, Y. (2017).

Artificial intelligence in healthcare: Past, present and future. Stroke and Vascular Neurology, 2(4), 230–

243. https://doi.org/10.1136/svn-2017-000101

5. van Breugel, M., Verschueren, C., Zuidmeer-Jongejan, L., & van Reijsen, F. (2023). Current state and

prospects

https://doi.org/10.1111/all.15849

6. World Health Organization. (2020). Risk assessment of food allergens: Part 1 – Review and validation

of the Codex priority allergen list through risk assessment.

https://www.who.int/publications/i/item/9789240042391

of

artificial

intelligence

in

allergy.

Allergy,

78(9),

2611–2623.

7. Jaiswal, Y. S., & Williams, L. L. (2025). The rising incidence of food allergies and infant food allergies.

Annual Review of Food Science and Technology, 16, 269–287. https://doi.org/10.1146/annurev-food-

111523-121952

8. Yang, X., Zhang, W., & Huang, Y. (2025). Novel artificial intelligence approach for nsLTP early

detection using NIRs data. Food Analytical Methods, 18(5), 1051–1063. https://doi.org/10.1007/s12161-

025-02851-6

9. Li, J., Kim, H., & Park, S. (2024). Point-of-care fluorescence biosensing system for rapid multi-allergen

screening. Sensors, 24(11), 3280. https://doi.org/10.3390/s24113280

10. Chen, Y., Gao, H., & Zhang, T. (2025). Deep learning for microbiome-informed precision nutrition.

National Science Review, 12(8), nwaf148. https://doi.org/10.1093/nsr/nwaf148

11. Mandracchia, F., Llauradó, E., Tarro, L., & Delgado-Noguera, M. F. (2020). Mobile phone apps for food

allergies or intolerances in app stores: Systematic search and assessment using the Mobile App Rating

Scale (MARS). JMIR mHealth and uHealth, 8(9), e18339. https://doi.org/10.2196/18339

12. Revelou, P. K., Zioga, E., & Kordas, G. (2025). Applications of machine learning in food safety and

HACCP Monitoring of Animal-Source Foods. Foods, 14(6), 922. https://doi.org/10.3390/foods14060922

13. Sarlakifar, F., Mehdizadeh, A., Rezaei, M., & Zolghadr, M. (2025). AllerTrans: A deep learning method

for predicting the potential allergenicity of protein sequences. BioMethods, 10(1), bpaf040.

https://doi.org/10.1093/biomethods/bpaf040

14. Liu, B., Yang, Z., Liu, Q., Zhang, Y., Ding, H., Lai, H., & Li, Q. (2023). Computational prediction of

Page 6799

www.rsisinternational.org

INTERNATIONAL JOURNAL OF RESEARCH AND INNOVATION IN SOCIAL SCIENCE (IJRISS)

ISSN No. 2454-6186 | DOI: 10.47772/IJRISS | Volume IX Issue X October 2025

allergenic proteins based on multi-feature fusion. Frontiers in Genetics, 14, 1294159.

https://doi.org/10.3389/fgene.2023.1294159

15. Kim, J. H., Koo, J. Y., Kim, J. S., & Om, A. S. (2025). Role of artificial intelligence (AI) and machine

learning (ML) in food safety and quality improvement. Food Supplements & Biomaterials Health, 5(2),

e11. https://doi.org/10.52361/fsbh.2025.5.e11

16. Konstantakopoulos, F. S., Georga, E. I., & Fotiadis, D. I. (2024). A review of image-based food

recognition and volume estimation artificial intelligence systems. IEEE Reviews in Biomedical

Engineering, 17, 136–152. https://doi.org/10.1109/RBME.2023.3283149

17. Bu, L., Hu, C., & Zhang, X. (2024). Recognition of food images based on transfer learning and ensemble

learning. PLOS ONE, 19(1), e0296789. https://doi.org/10.1371/journal.pone.0296789

18. Rita, L., Southern, J., Laponogov, I., Higgins, K., & Veselkov, K. (2024). Optimizing ingredient

substitution using large language models to enhance phytochemical content in recipes. Machine Learning

and Knowledge Extraction, 6(4), 2738–2752. https://doi.org/10.3390/make6040131

19. Pati, S., et al. (2024). Privacy preservation for federated learning in health care. Journal / Repository.

Retrieved from https://www.ncbi.nlm.nih.gov/pmc/articles/PMC11284498/

20. Zheng, J. (2024). Artificial intelligence applications to measure food and nutrient intakes: A scoping

review. JMIR (Journal of Medical Internet Research), 2024. https://www.jmir.org/2024/1/e54557/

Page 6800

www.rsisinternational.org