A Review: Data Overfitting and Underfitting Techniques

Modern problem statements demand advanced, futuristic, cutting-edge, progressive and trend-setting solutions. In today’s landscape, Artificial Intelligence (AI) stands out as a revolutionary, data-driven technology reshaping our perception of modern applications. However, when AI interacts with Machine Learning (ML), another data-driven domain, a significant hurdle emerges concerning the quality and quantity of data. This problem profoundly impacts the performance of ML Models, prompting engineers and architects to address a myriad of questions before selecting a suitable mode. Consequently, altering the architecture and models becomes arduous due to data-related issues, which can be numerous. Hence, a solution must be sought that caters to both needs. This review paper delves into the various challenges related to Data Overfitting and Data Underfitting, exploring methods for identification, key questions for consideration, underlying reasons, techniques, real-life examples, synthetic data illustrations, and a plethora of case studies from diverse research papers showcasing how they tackled the problem and how overfitting and underfitting can be solved.