Mathematics Comprehensive Review of Machine Learning Methods and Applications in Economics

The last few years has seen great advancements in both the availability of data and the availability of computing power. This has made it possible to integrate many new machine learning techniques into economic studies. Even though traditional econometric approaches are used to formalize economic problems and help with predictions, they usually collapse under their own complexity and the estimation of the given model’s parameters. Likewise, the statistical inferences made are typically misappropriated. This is because they are based on multiple restrictive assumptions. In contrast, the machine learning approach is far more robust and seeks to provide the best possible prediction irrespective of the guideline/model used. In this paper, we will provide a full, comprehensive, and in-depth study of the available literature associated with machine learning (ML), deep learning (DL), reinforcement learning (RL), and deep reinforcement learning (DRL), and their associations with economics. We will define the algorithms associated with each of these concepts, and then evaluate the literature associated with each algorithm concerning core economic concepts, such as economic forecasting, the economics of financial markets, the economics of public policy, and economics in the presence of uncertainty. We will then present the results of a systematic (PRISMA) based literature review in which we evaluated peer-reviewed literature to uncover the main methodological and substantive foci, as well as uncover the gaps in the literature. The analytical framework focused on deep reinforcement learning (DRL), as it is the form of machine learning that most closely approximates the ability of multiple traditional economic models to solve complex dynamic problems. The paper also proposes a generalized framework for applying DRL in economic analysis and discusses key limitations related to data requirements, interpretability, and stability. By bridging mathematical theory and applied economics, this review aims to provide researchers and practitioners with a structured reference for selecting, implementing, and extending modern machine learning methods in economic research.