Combining Machine Learning And Column Generation Techniques To Solve The Set Cover Model For A Multiple Container Loading Problem

In many practical scenarios,choices of containers of different sizes and costs available,and the task is to select a set of containers that can hold all the goods while minimizing the cost.This problem is known as the multiple container loading cost minimization problem(MCLCMP).Many hard combinatorial optimization problems including the multiple container loading problem(MCLP)can be formulated as a set cover model.Due to its complexity,the decision process is often divided into two stages,the first stage ignores the sub-decision details and roughly estimates the global decision;the second stage consider all the details and generate a complete solution.Such problems are usually solved through column generation(CG)techniques.The effectiveness of CG largely depends on how efficiently the pricing sub problems can be solved.Our strategy: 1.relax the pricing sub problem.2.train a machine learning model to predict the difference between a relaxation and its true solution.3.turn a relaxation into a true solution based on prediction.We know that each company has different product specifications and unique operational data,machine learning can mine the information behind the data,so we improve the effectiveness of CG for the MCLP by extracting valuable information from the accumulated historical operational data.Thereby allocating more time for the global decision,and make it more quickly lock the high-quality solution space,and ultimately improve the overall performance of the optimization algorithm.At the same time,the improved optimization algorithm can achieve higher performance when applied to a specific enterprise,this is the traditional optimization algorithm cannot do?...