Classification Of GEP-based Multi-gene Family Encodes A Complex Network Of Associations

With the rapid development of information technology, there are all kinds of large scalenetwork in human life. It is very necessary to make the research of complex networks for its widespread. And the researches of complex network mainly focus on the following four aspects. Theyare the statistical characteristics and analysis of complex network topology, complex networkformation mechanism and evolution model, the dynamics research of complex network, andcomplex community network. It is very helpful for understanding the structure and characteristicsof the whole network to detect and divide the complex community network. And it can guidepeople to make decisions，has very strong practical signiifcance.At present, many scholars have proposed a variety of complex community network divisionalgorithms, and obtained certain achievement. However, they have some shortcomings, such aslow searching accuracy, high time complexity and needed to know in advance the numbers andnodes of the community network. Therefore it is not suitable for the analysis of large complexnetwork. In view of the above, a new complex community network division algorithm based onmulti-gene families (MGF) is proposed in this paper. And the experiment and analysis show thatthis algorithm is more eiffcient and accurate than the traditional evolutionary algorithms.The major work and innovations of this paper are as followings:(1)A new encoding method, named multi-gene families (MGFs) encoding，is proposed toapply for complex community network. In this new encoding method, it takes advantage of theMGF’s characteristic in gene expression programming to respectively encode the node ID and thecommunity type into the same chromosomes of two different MGFs, and the serial encodingmethod overcomes the shortcomings of traditional tree encoding method.(2)A new complex community network division algorithm based on multi-gene familiesencoding is proposed in this paper. It takes advantage of the combination of optimization specialoperators in gene expression programming, such as inversion, exchange restrictions, generalizedexchange, gene deletion/insertion and gene segments deletion/insertion. And it is more eiffcientthan the classical genetic algorithm. In this paper, the operators of inversion and exchangerestrictions are introduced to promote the algorithm convergence and reduce the time complexityeffectively. Meanwhile, the experiment and analysis show that this algorithm is more eiffcientand accurate than the traditional evolutionary algorithms.(3)A new elite migration operator is imported in the evolutionary stage of gene expressionprogramming. The elite migration strategy is applied to the whole hereditary stage to speed up theconvergence and improve accuracy, such as gene selection, chromosome crossover, chromosome inversion, and exchange restrictions. Namely, we can ifnd out the best individual in the geneticoperator before and after modiifcation for all chromosomes of a generation respectively. If thebest individual in front of the modiifed is better than the individual after modiifcation, let the bestindividual pre-modiifed replace the worst individual in the population of the modiifed. Using elitemigration strategy can not only control the genetic evolutionary direction, but also improveaverage iftness of per generation.