Deep Study on the Application of Machine Learning in Bin Packing Problems

Abstract

This paper conducts a comprehensive review of literature focusing on strategies applied in the realm of Machine Learning (ML) in a period from ten years ago to the present to address the Bin Packing Problem (BPP) and its various variants. The Bin Packing Problem, a renowned optimization challenge, involves efficiently allocating items of varying sizes into containers of fixed capacity to minimize the number of containers used. Despite the extensive body of research and the existence of heuristic algorithms, unresolved challenges persist in BPP's solution. This deep study systematically explores the landscape of ML applications, delving into innovative approaches and methodologies proposed for tackling BPP and its diverse extensions, including 2D-BPP, 3D-BPP, Multi-objective BPP, and dynamic variants. The review critically examines the performance and contributions of ML-based strategies, shedding light on their efficacy in optimizing the packing process. Key findings highlight the promising directions taken by ML in solving complex optimization problems, emphasizing its potential to enhance BPP solution methodologies. The synthesis of diverse ML strategies and their integration with traditional heuristics forms a central theme, showcasing the evolving landscape of research in this domain. Additionally, this review identifies gaps and future research directions, emphasizing the relevance and effectiveness of ML as a valuable tool for improving performance in resolving BPP and its related challenges. The insights derived from this study aim to guide researchers, practitioners, and decision-makers in understanding the current state of ML applications in the context of Bin Packing Problems and inspire further advancements in this field.