Research On The Improvement And Application Of Marine Predator Algorith

Marine Predator Algorithm(MPA)is an emerging swarm intelligence optimization algorithm,which has good performance in solving various practical optimization problems,but there are also problems such as insufficient optimization accuracy,single search path,and slow convergence speed.In this paper,the algorithm of marine predator is improved by introducing individual renewal strategy and individual disturbance strategy to improve its convergence accuracy and convergence speed,the improved marine predator algorithm is then used to optimize the K-Nearest Neighbor(KNN),and the optimized KNN is used for the differential diagnosis of erythematous squamous disease.The specific work is as follows:(1)Aiming at the problem of insufficient optimization ability of marine predator algorithm in solving complex optimization problems,it is proposed to use the emission iterative strategy to complement the iterative strategy of the original MPA to improve the diversity of particle motion paths and strengthen the search of feasible domains by particles.At the same time,a sine cosine perturbation strategy is added to perturb the particles and increase the local development ability of the particles.The improved marine predator algorithm of the above two strategies is called OMPA,which tests OMPA with MPA,MPASCA,MPAOBL,PSO,SCA,SSA,GWO in 10 single-peak,7 multi-peak and6 fixed-dimensional typical optimization problems,and compares each algorithm from multiple aspects such as complexity,convergence accuracy,convergence speed,running time and performance on high-dimensional optimization problems.The results show that OMPA has better optimization accuracy and convergence speed than the above algorithm without increasing the time complexity of the original algorithm,and also performs well in high-dimensional optimization problems.(2)Aiming at the shortcomings of KNN in high-dimensional cases with large calculation amount and low classification accuracy,the improved marine predator algorithm is used for both feature selection and parameter optimization of KNN to improve the classification effect of KNN(OMPA-FSPO-KNN),and the model is used for the diagnosis of erythematous squamous disease.OMPA-FSPO-KNN is compared with OMPA-FSKNN,MPA-FSPO-KNN,KNN and SVM in three evaluation indexes: accuracy,Kappa value and AUC value.The results show that the OMPA-FSPO-KNN model is the best among the five models in the three evaluation indexes,with an accuracy of 98.61%,a Kappa value of 0.9821 and an AUC value of 0.9923,and the OMPA-FSPO-KNN has high classification accuracy and good stability,which could make effective diagnosis and identification of erythematous squamous disease.Finally,the OMPA-FSPO-KNN is used to extract the important feature variables affecting the diagnosis of erythematous squamous disease,and 12 important feature variables were extracted for decision-making reference.