Research On Swarm Intelligence Optimization Algorithms Based On Multiple Swarms And Their Medical Diagnosis Applications

After human society entered the network era,with the continuous development of information technology,the amount of data generated and collected in various production activities has continued to increase,prompting a whole new transformation in the field of machine learning and data mining research.Swarm intelligence optimization algorithms are an important research direction in the field of evolutionary computing.In recent years,there are many excellent swarm intelligence optimization algorithms emerged.Among them,the Sine Cosine Algorithm(SCA),Harris Hawks Optimization(HHO),and Salp Swarm Algorithm(SSA),all of which have simple structures and high performance,have been widely used in the field of optimization problems.They are widely used in the field of optimization problems.However,their original algorithms are prone to local optimality and slow convergence when dealing with complex optimization problems,and there is room for improvement.The focus of this thesis is to improve these three types of group intelligence optimization algorithms and use their optimization performance to handle feature selection applications for medical datasets.Distinguishing the strength of relevance of features,removing useless features,and improving the quality of raw datasets is a key focus in the field of machine learning and data mining research.The specific operation of feature selection is to find the best set of features using established optimization algorithms.The main objective is to find the set of features with the strongest relevance to the prediction results.In this way,it is possible to improve the efficiency of processing the dataset,speed up the training of the machine learning model and finally improve the classification ability of the overall combined model.In this thesis,an improved swarm intelligence optimization algorithm based on multiple swarm mechanisms is used to solve the feature selection problem for medical datasets and provide practical application help for medical diagnosis.In this thesis,the original algorithms of the bottle sea-sheath optimization algorithm,the positive cosine optimization algorithm,and the Harris eagle optimization algorithm are improved by introducing various efficient search mechanisms to redesign the population structure and improve the population diversity using multiple swarm mechanisms.After that,relevant feature selection schemes are developed for the improved algorithms.Finally,they were combined with specific classifier algorithms for application in the field of medical disease diagnosis.The main research focus of this thesis is as follows.(1)To address the weaknesses of the original basic optimization algorithm of slow convergence and inability to jump out of the local optimum,we propose three improved swarm intelligence optimization algorithms based on multiple swarm mechanisms on the basis of the original positive cosine optimization algorithm,Harris hawks optimization algorithm,and bottle sea squirt optimization algorithm.The introduction of the application of multiple swarm mechanism can effectively protect the diversity of the algorithm population during the iterative process of the algorithm to deal with the optimization problem.In addition,this thesis also incorporates a variety of other effective algorithm improvement mechanisms into the improved optimization algorithms to enhance the local search and global search capabilities.(2)The second research focus of this paper lies in testing the performance of the improved swarm intelligence optimization algorithm.In this regard,we test the performance of the improved algorithm with different benchmark function test sets and conduct comparison experiments with other excellent intelligent optimization algorithms.The experimental results demonstrate that the improved optimization algorithm performs well on the benchmark function test set and achieves a good balance between global and local search.It can be extended and applied to various optimization fields to solve optimization problems.(3)The third important piece of research in this thesis lies in the application of a variety of swarm-improved swarm intelligence algorithms that have been obtained to the feature selection problem in medical diagnosis.The improved algorithms are combined with specific classifier algorithms to perform experiments on feature selection on specific medical datasets with a view to finding the feature subset and optimization algorithms with optimal results.Different improvement mechanisms are used to enhance the specific performance of each type of algorithm in optimizing the feature set and finally obtaining discrete solutions representing the selection of specific feature sets.With the KNN(K Nearest Neighbors,KNN)classifier algorithm,the fitness value of each solution in the population of the optimization algorithm is evaluated using the number of features selected and the classification error rate in the dataset.The specific dataset applied is 15 medical datasets of varying dimensions taken from the UCI(University of California,Irvine)machine learning public dataset repository.The experimental results found the optimal combinatorial model among the three types of algorithms,and also verified that the combinatorial model proposed in this thesis is efficient in improving the classification accuracy and screening the set of valid features,and can be effectively applied in the field of medical diagnosis.