| Swarm intelligence optimization algorithms are an efficient class of algorithms inspired by the collective behavior of organisms in nature.Its population consists of relatively simple individuals,but can exhibit coordinated self-organization behavior based on easy interaction rules and eventually converge to the global optimal solution.In recent years,machine learning has developed rapidly and been widely used in various research fields,among which the combination of swarm intelligence algorithms and models in machine learning has attracted the interest of researchers.This combination can greatly optimize and improve the effectiveness of machine learning models.Therefore,two swarm intelligence algorithms are studied in this paper,namely firefly algorithm and black hole algorithm.These algorithms have their own advantages and disadvantages,and they have mutual complementary reference value in the improvement of the algorithms.Therefore,this paper improves the performance of these two algorithms by designing novel improved strategies and applies the proposed algorithms to solve optimization problems related to improving the performance of machine learning models.The main contributions of this paper are shown below:(1)An improved firefly algorithm with the Relief F-based initialization method and the weighted voting mechanism is proposed and applied to the feature selection problem.Firstly,a feature grouping initialization method combining the results of Relief F algorithm and cosine similarity is designed to replace random initialization.Then,the direction of the firefly is modified to move toward the current global optimal solution.Finally,a weighted voter inspired by the ensemble algorithms is proposed to build the recommended positions of fireflies,which is also combined with the elite crossover operator and the mutation operator to improve the diversity of the population.Simulation experiments based on standard datasets verify superiority of this algorithm compared to the selected state-of-theart comparison algorithms in feature selection problem.(2)An improved self-adaptive logarithmic spiral path black hole algorithm is proposed and applied to the selection of the initial centroids for K-Means algorithm.Firstly,the logarithmic spiral path and random vector path are introduced into the black hole algorithm.Meanwhile,a parameter is used to control the randomness of path selection,which enhances the local exploitation capability of the algorithm.In addition,a replacement mechanism of stars is designed and an adaptive parameter is introduced to control the replacement mechanism,which helps maintain a balance between the exploration and exploitation capabilities of the algorithm.Finally,the simulation experimental results show that the proposed algorithm performs better than other comparative algorithms in the selection of the initial centroids for K-Means algorithm.(3)A hybrid firefly and black hole algorithm for parameter tuning of XGBoost method is proposed and applied to the study of intrusion detection systems.Firstly,the algorithm designs a double black hole mechanism by introducing the concept of the second black hole and adjusting the moving trajectory of the stars using the attraction of both black holes.Secondly,an improved initialization method of the stars is proposed,where a star that crosses the event horizon of the black hole has an opportunity to be replaced by a new star around the black hole.Finally,a combination of the firefly perturbation strategy and mutation operator is proposed to improve the global search capability of the algorithm.The effectiveness of the proposed method on the XGBoost parameter tuning problem and the feasibility of the strategy in intrusion detection applications are verified through comparison experiments based on the NSL-KDD dataset. |
