Research On Improved Electric Fish Optimization And Its Application

The Electric Fish Optimization Algorithm(EFO)is a new meta-heuristic algorithm.It was inspired by the prey location and communication behaviour of electric fish.It has attracted the attention of researchers because of its simplicity and easy implementation.However,due to the influence of the optimization mechanism,it is still difficult to avoid some shortcomings,such as: precocious convergence,exploration and mining imbalance,and local optimization.This makes it increasingly difficult for EFO to solve complex optimization problems.However,with the development of science and technology,the dimension of practical optimization problem is increasing day by day,many researchers have begun to use the more robust group intelligence optimization algorithms to solve such optimization issues.Therefore,in order to improve the convergence performance of EFO and balance its search and mining capabilities,this thesis does the following work:(1)In order to overcome the shortcomings of the EFO algorithm and improve the performance of the EFO algorithm,an orthogonal fish optimization algorithm with quantification(QOXEFO),is proposed.First,orthogonal cross-design and quantification technique are employed to enhance the diversity of population and convergence precision of EFO.Second,a dynamic boundaries mechanism is adopted to boost the convergence speed.Third,a sinusoidal-based active electrical positioning update strategy is used to change the direction of movement of an individual,thereby helping them jump out of local optimality.Finally,in order to confirm the performance of the proposed QOXEFO algorithm,the CEC2017 benchmark functions are utilized by comparing with other 8well-known evolutionary algorithms of solution accuracy and convergence speed.Experimental and statistical analysis results show that QOXEFO has the promising performance.(2)In order to improve the ability to solve feature selection problems,this thesis proposes a binary quantization orthogonal crossover electric fish optimization(BQOXEFO)feature selection method.In the initialization stage,the transformation function is used to binary the independent variables,and the fitness function is given according to the accuracy of the classification and the number of selected features,and the data is classified using the K-Nearest Neighbour(KNN).To verify the performance of the BQOXEFO algorithm,15 benchmark datasets in the UCI database are utilized by comparing with 8classical algorithms of the solution accuracy and convergence speed.Statistical analysis results show that BQOXEFO has the promising performance,it can well complete the feature classification task,so that the number of features is minimized and the classification performance is maximized.