Generative-Agentic AI (GA-AI) Framework for Product Recommendation

Gonugunta, Durga Nagendra Prasad (2025) Generative-Agentic AI (GA-AI) Framework for Product Recommendation. Masters thesis, Dublin, National College of Ireland.

Preview	PDF (Master of Science) Download (659kB) | Preview
Preview	PDF (Configuration Manual) Download (290kB) | Preview

Abstract

Availability of millions of products online has made the product recommendation systems crucial. Product recommendation systems suggest relevant products to the users based on their requirement. Current research in product recommendations rely on Collaborative Filtering and Deep Learning models but these lack conversational ability. Building a product recommendation system that understands user intent, retrieves the product and generates conversational response is a challenge. This research proposes a Generative-Agentic AI framework to deliver product recommendations through conversation. The proposed framework combines Intent Classification Agent, Retrieval Agent, Generation Agent and Controller Agent for product recommendation through conversation. Amazon 5 core product review dataset covering six categories of Toys and Games, Musical Instruments, Cellphones and Accessories, Appliances, All Beauty and Amazon Fashion has been used for this research. Data Analysis techniques have been applied on this dataset to build Embeddings from it. Logistic Regression for Intent classification, Facebook AI Similarity Search (FAISS) index with MiniLM embeddings for Retrieval Agent and Quantized OpenHermes Language Model for Generation Agent are combined to build an Agent based recommendation framework. Results are evaluated using classification accuracy for Intent Classification Agent, Mean Reciprocal Rank, Hit Rate and Relevancy score for Retrieval agent and BLEU, BERTScore, ROUGE-L score, faithfulness and relevancy for Generation Agent. The research shows promise for a conversation based Product Recommendation using Generative Agentic AI Framework. It benefits the community by demonstrating competitive accuracy and reliability, while reducing computational cost.

Item Type:	Thesis (Masters)
Supervisors:	Name Email Stynes, Paul UNSPECIFIED
Subjects:	Q Science > QH Natural history > QH301 Biology > Methods of research. Technique. Experimental biology > Data processing. Bioinformatics > Artificial intelligence Q Science > Q Science (General) > Self-organizing systems. Conscious automata > Artificial intelligence P Language and Literature > P Philology. Linguistics > Computational linguistics. Natural language processing
Divisions:	School of Computing > Master of Science in Artificial Intelligence
Depositing User:	Ciara O'Brien
Date Deposited:	28 May 2026 13:54
Last Modified:	28 May 2026 13:54
URI:	https://norma.ncirl.ie/id/eprint/9320

Actions (login required)

View Item