SVM-Assisted Decomposition-Based Multi-objective Evolutionary Algorithm Research And Application

Multi-objective evolutionary algorithm based on decomposition(MOEA/D)has become an effective method for solving multi-objective optimization problems and has been widely used.Although the performance of MOEA/D is outstanding in solving most multi-objective optimization problems,there is still some room to be explored to further improve the diversity and convergence of the algorithm and reduce the computational cost in engineering applications.The research of this thesis includes the following two aspects.(1)After extensive experiments,researchers found that the choice of the propagation operator in MOEA/D has a great impact on the quality of the solution,and a suitable operator can improve the performance of the algorithm to a great extent.Therefore,based on MOEA/D,this thesis proposes a hybrid operator selection strategy(HOSS),which integrates HOSS into MOEA/D(called MOEA/D-HOSS)to improve the performance of the algorithm.MOEA/D-HOSS applies SVM classification prediction and effect feedback Compared with the traditional MOEA/D,which uses only a single multiplication operator,MOEA/D-HOSS uses multiple operators to participate in the evolutionary search for more diverse and better offspring.Also,using the trained classifier to do pre-selection of the generated offspring individuals reduces the number of evaluations and speeds up the convergence of the algorithm.Finally,MOEA/D-HOSS was experimented on multiple sets of multi-objective test problems and analyzed in comparison with some other current excellent multi-objective evolutionary algorithms on different evaluation metrics,and the experiments verified the effectiveness of MOEA/D-HOSS.(2)A dual-objective warehouse space dynamic planning system is designed to improve the operational efficiency of goods and warehouse space utilization during warehouse storage,and at the same time,warehouse information can be captured and shared in real time.The system takes the highest utilization rate of warehouse space and the shortest picking distance as the optimization objectives,and adopts a decomposition-based multi-objective evolutionary algorithm with a user preference mechanism in order to better balance these two optimization objectives and obtain a superior solution.The system not only realizes the basic functions of order inbound and outbound,querying,getting the current status of the warehouse and adding supplier information,but also realizes the function of selecting the optimal warehouse location for new orders by order information.The experimental results show that the system can effectively recommend a reasonable warehouse location.