Research On Clustering Algorithm Of Immune Regulation Network

Data mining is a basic step in the knowledge discovery process in the databases.It is a process of extracting useful information and knowledge from rich,incomplete,fuzzy and random data,and clustering is an unsupervised machine learning method in data mining,which can analyze the dataset without the class.With the rapid development of information technology,the scale of data is more and more large,the data type is becoming more and more diversified.The traditional clustering algorithm can't solve the current problem effectively.This thesis will improve the artificial immune network clustering algorithm.The immune regulation network clustering algorithm based on the Manifold distance is proposed to improve the clustering accuracy.And the algorithm combines the BIRCH algorithm and the quantum immune mechanism which can effectively deal with the large-scale datasets.The specific work of this thesis is as follows:(1)Firstly,we propose an immune regulation network clustering algorithm based on the Manifold distance.The stimulation regulation factor is introduced to replace the clone operator and mutation operator in artificial immune network,antibodies with high stimulation regulation factor are selected by the immune regulation mechanism,and calculate the affinity by Manifold distance.The proposed algorithm not only overcomes the shortcoming that the artificial immune network algorithm is sensitive to noise,but also overcomes the shortcoming that the improved artificial immune network clustering algorithm obtained poor results on the Manifold-distribution datasets and real datasets.In the comparative experiment,the algorithm proposed in this thesis obtains a good clustering result on the artificial datasets and the UCI datasets.It is proved that the proposed algorithm has better clustering performance.(2)Aiming at the problem that the immune regulation network algorithm can't deal with large-scale datasets,a large-scale data clustering algorithm based on immune regulation network is proposed.This algorithm combines BIRCH algorithm and immune regulation network clustering algorithm.The algorithm classifies the original dataset,divides the large original dataset into subclasses with different number of sample points,extracts the centers of the subclasses,and sets these smaller clustering centers as a new dataset,then uses the immune regulation network clustering algorithm to cluster.This combination not only solves the problem that the immune regulation network is not suitable for large-scale datasets because of the high computational complexity,but also solves the problem that the BIRCH algorithm needs the number of clusters and can't deal well with complex real datasets.In the comparison experiment we can see that our proposed algorithm has better clustering performance.(3)Finally,in order to reduce the parameter influence of proposed algorithm and improve the clustering accuracy of the algorithm,a large-scale data clustering algorithm based on quantum artificial immune network is proposed.When the artificial immune network is improved,the quantum clone,quantum mutation,and the quantum crossover operator are introduced,so that the algorithm avoids the local optimal solution and premature phenomenon.By improving the BIRCH algorithm,it can adaptively obtain the natural death threshold and reduce the sensitivity of the algorithm to the parameter.Then,compared with the large-scale data clustering algorithm based on immune regulation network,it is verified that the proposed algorithm based on quantum artificial immune network not only can get better clustering results,but also greatly reduce the influence of algorithm parameters.