Study On Swarm Intelligence And Its Applications To Radar Signal Classification

Radar signal classification is an important issue in electronic reconnaissance, it is the process of isolating every radar pulse sequence from the random overlapping pulse stream. This kind of classification is realized by making use of the signal parameter relevance of the same radar and between different radar. This paper has studied the swarm intelligence based data mining method and its applications to radar signal classification.In this paper, combined with the purpose and the requirements of the radar signal classification, the feature selection and the clustering problems in data mining field have been regarded as a single objective optimization problem, they have been applied to solve the feature selection and radar signal classification problem. From the angle of particle swarm optimization algorithm, a kind of discrete particle swarm optimization algorithm is used to solve the two problems. In this paper, the main work is as follows:(1) Introduce the basic ABC’s of data mining, with emphasis on feature selection and clustering. Introduce the theory of swarm intelligence including ant colony optimization and particle swarm optimization with emphasis on the basic theory, framework and variants of particle swarm optimization.(2) This paper has put forward a local learning based discrete particle swarm optimization method for data feature selection. The particle swarm optimization algorithm is widely used for solving continuous optimization problems, however, in reality many problems are discrete, i.e., the particle is integer coded, to design discrete particle swarm optimization techniques has aroused scholars’interest. This paper has proposed a discrete particle’s status and update principles have been redefined, a forward floating based greedy local search is designed. Experiments on10UCI data compared with3classical feature selection methods demonstrate that the proposed algorithm is effective.(3) This paper has proposed an adaptive discrete particle swarm optimization based clustering method. Conventional clustering methods cannot determine the clusters adaptively. For the very clustering problem, in the proposed method, particle has been redefined under discrete context, particle’s status has been reconsidered. In order to improve its searching ability, a greedy local learning strategy is designed. In order to check the performance of the proposed algorithm, extensive experiments on the UCI ata sets have been done, it has been compared with several classical clustering methods also. Experiments indicate that the proposed method has good clustering performance, it can solve the radar signal clustering problem effectively.