Enhancing decision making through similarity-driven knowledge integration in resource allocation and content matching.

This research aims to build a novel framework that enhances decision-making through an integration of similarity driven Case-Based Reasoning (CBR) with advanced Large Language Model (LLM) techniques via Retrieval Augmented Generation (RAG) and Genetic Algorithm (GA) optimisation. Currently, experimental work focuses on refining the loss function components to tune angle-optimised embedding models using both semi-supervised and unsupervised approaches. In parallel, experiments are being conducted to fine-tune LLMs as baselines for evaluation and to determine the best way to use LLMs as evaluative judges. Preliminary data analysis and enrichment have been conducted on operational datasets (e.g., WM Nicol company records). The final goal is to advance the state-of-the-art in CBR methods while providing a robust foundation for adaptive, context-aware decision support across multiple domains.