Towards Safe and Personalized Skincare Recommendations: Integrating CNN-Based Skin Analysis, Retrieval-Augmented Generation, and Allergy-Aware Guardrails
Authors
Assistant Professor, Department of Computer science, Shri Shankarlal Sundarbai Shasun Jain College for Women, Chennai, India (India)
Student, Department of Computer Science Shri Shankarlal Sundarbai Shasun Jain College for Women, Chennai, India (India)
Student, Department of Computer Science Shri Shankarlal Sundarbai Shasun Jain College for Women, Chennai, India (India)
Article Information
DOI: 10.51583/IJLTEMAS.2026.151000070
Subject Category: Artificial Intelligence
Volume/Issue: 15/10 | Page No: 840-845
Publication Timeline
Submitted: 2026-10-22
Accepted: 2026-10-27
Published: 2026-11-07
Abstract
Skincare has become a routine, high frequency purchasing decision for consumers worldwide, yet most buyers select products with limited knowledge of ingredient composition or compatibility with their own skin, leading to financial waste, allergic reactions, and environmental waste. Existing AI driven skincare systems address only part of this problem: CNNs reliably classify skin type from a photograph, but the recommendation step typically relies on collaborative filtering or static rule based matching, which fails for new users and cannot easily incorporate new products. None of these systems verify whether a recommendation is actually safe for a user's allergies or sensitivities. This paper proposes a hybrid architecture combining CNN based skin classification, retrieval augmented generation (RAG) for live ingredient data, a deterministic rule based safety filter, and an LLM for personalized explanation each assigned a distinct role rather than one model handling everything. This design reduces the risk of unsafe recommendations while retaining up to date, personalized guidance. Empirical implementation and testing are identified as future work.
Keywords
skincare recommendation, convolutional neural network, retrieval augmented generation, large language model, guardrail architecture, allergy safety, cold start problem
This article is licensed under a Creative Commons Attribution 4.0 International License (CC BY 4.0).
Downloads
References
1. Fortune Business Insights, "Skincare Market Size, Share, Trends | Growth Analysis [2034]," 2026. [Online]. Available: https://www.fortunebusinessinsights.com/skin-care-market-102544 [Google Scholar] [Crossref]
2. "Human skin type classification using image processing and deep learning approaches," ScienceDirect, 2023. [Online]. Available: https://www.sciencedirect.com/science/article/pii/S2405844023083846 [Google Scholar] [Crossref]
3. V. [author], "A Recommendation System for Facial Skin Care using Machine Learning Models," GitHub / IEEE. [Online]. Available: https://github.com/vinit714/A-Recommendation-system-for-Facial-Skin-Care-using-Machine-Learning-Models [Google Scholar] [Crossref]
4. "Enhanced CNN Architecture for Multi-label Skin Condition Classification for Personalized Skincare Recommendations," in Proc. Springer Nature Link, 2026. [Google Scholar] [Crossref]
5. "Adaptive Candidate Retrieval with Dynamic Knowledge Graph Construction for Cold-Start Recommendation," arXiv:2505.20773, 2025. [Google Scholar] [Crossref]
6. "RAMO: Retrieval-Augmented Generation for Enhancing MOOCs Recommendations," arXiv:2407.04925, 2024. [Google Scholar] [Crossref]
7. H.-D. Kieu and T.-S. Nguyen, "Keyword-driven Retrieval-Augmented Large Language Models for Cold-start User Recommendations (KALM4Rec)," arXiv:2405.19612, 2024. [Google Scholar] [Crossref]
8. "ReRag: A New Architecture for Reducing the Hallucination by Retrieval-Augmented Generation," in Proc. IEEE Conf., 2025. [Online]. Available: https://ieeexplore.ieee.org/document/10773428/ [Google Scholar] [Crossref]
9. "LLM-Based Guardrail Models," EmergentMind, 2025. [Online]. Available: https://www.emergentmind.com/topics/llm-based-guardrail-models [Google Scholar] [Crossref]
Metrics
Views & Downloads
Similar Articles
- Impact of AI Tools on Student Learning
- Repogent: An Autonomous Multi-Agent System for End-To-End Repository Maintenance
- Trust-Centric Federated Learning Using Blockchain-Enabled Smart Contracts: An Analysis of Data Privacy, Model Integrity, and Decentralized AI Governance
- Can AI Care? Rethinking Ethics, Clinical Integrity, and Professional Responsibility in the Digital Age of Mental Healthcare
- AI-Based Smart Irrigation System Using IoT and Machine Learning for Precision Agriculture
This work is licensed under CC BY 4.0