Prescriptive Analytics: Review of Frameworks and Critical Evaluation of PrescrX

The thesis titled ``Prescriptive Analytics: Review of Frameworks and Critical Evaluation of PrescrX'' explores the landscape of decision-making systems powered by data-driven prescriptive techniques. After reviewing existing frameworks, the work introduces PrescrX---a custom system developed to tackle practical optimization problems based on predictive input. The author outlines how prescriptive analytics bridges the gap between analytical insight and actionable strategy. The literature review delves into well-established models including optimization, simulation, and decision analysis, with particular attention to the integration of machine learning and mathematical programming. The PrescrX tool is then analyzed in depth, both in terms of its operational logic and its architectural design. To objectively evaluate the performance of PrescrX and support comparisons with traditional optimizers, the thesis proposes a set of original metrics aimed at assessing the quality of the generated prescriptions. These metrics are designed to be interpretable, reproducible, and applicable across multiple case studies. To validate the model in practice, two distinct case studies are employed: the well-known MNIST dataset, representing a high-dimensional image classification problem, and a real-world industrial dataset involving binary classification of component cleanliness based on operational parameters. These case studies allow for the examination of both computational behavior and domain-specific applicability of the tool. The thesis concludes with a critical analysis of PrescrX’s performance and provides suggestions for future improvements and broader applications in industrial and research contexts.