Research On Intelligent Optimization Algorithm Based On Online Hybrid Landscape Analysis

With the development of social economy,people are facing more and more complex optimization problems,and the performance of intelligent optimization algorithms is facing great challenges.Although,many different types of improved algorithms have been proposed one after another,their performance still has certain shortcomings,such as slower convergence speed and poorer optimization-seeking accuracy.The main reason for such shortcomings is that many related works are based on the historical experience of algorithms for algorithm design,while ignoring the impact of problem fitness landscape features on algorithm performance.To this end,this paper studies the design of intelligent optimization algorithms from the perspective of problem fitness landscape features,and proposes an online hybrid landscape analysis technology by combining two offline fitness landscape analysis techniques: information landscape and information entropy,which can analyze the problem fitness landscape features to adjust the search behavior of the algorithm in real time during the execution of the algorithm,and strongly bridges the gap between algorithm design and problem fitness landscape features.The main work and contributions of this paper are as follows:(1)The proposed online hybrid landscape analysis technology is more efficient and has lower time complexity compared to existing related methods.To verify this,we compared it with two existing related works: the artificial bee colony algorithm based on online dispersion Metric technology,and the differential evolution algorithm based on fitness-distance correlation analysis technology,and designed a comparative algorithm based on online hybrid landscape analysis technology.Experiments were conducted on two sets of test functions,CEC 2013 and CEC 2014,and the results showed that the comparative algorithm had better solving performance and less running time than the original algorithm.(2)The proposed online hybrid landscape analysis technology can be used to improve multi-strategy mechanisms and is more efficient than multi-strategy mechanisms based on historical empirical information.To verify this,we took a representative work,the Multi-strategy ensemble artificial bee colony algorithm(MEABC),and designed a corresponding comparative algorithm: the multi-strategy artificial bee colony algorithm based on online hybrid landscape analysis technology,which modifies the multi-strategy mechanism of MEABC from the perspective of problem features.Experiments were conducted on two sets of test functions,CEC 2013 and CEC 2014,and the results showed that the comparative algorithm outperformed the original algorithm in terms of result accuracy.(3)The proposed online hybrid landscape analysis technology can be used to improve algorithm integration mechanisms and is more efficient than integration mechanisms based on historical empirical information.To verify this,we took a representative work,the ensemble of differential evolution variants(EDEV),and designed a corresponding comparative algorithm: the evolutionary differential evolution algorithm based on online hybrid landscape analysis technology,which modifies the algorithm integration mechanism of EDEV from the perspective of problem features.Experiments were conducted on two sets of test functions,CEC 2005 and CEC 2017,and the results showed that the comparative algorithm also outperformed the original algorithm in terms of result accuracy.