Robust Adaptive Graph-Regularization Clustering Algorithm

In the 21st century,Internet and data mining technologies are booming,and society is rapidly entering the era of big data.The continuous and growing big data has improved people's way of life,while its storage,mining and application have become a challenge for people in the information age.Obtaining the potential knowledge efficiently and low cost from massive data has become a major problem for scholars.In recent years,cluster analysis technology is highly valued in the field of pattern recognition and data mining.Traditional clustering methods such as k-means clustering algorithm and spectral clustering algorithm have been applied to various fields.However,the traditional graph-based clustering method is usually based on a given data graph and its adjacency graph.In clustering,we need an effective algorithm to get the result of data clustering.However,there are often noise in the data.The near neighbor graph can not meet the needs of practical application.To solve the above problems,scholars have proposed many algorithms to reduce the effect of error,noise and outliers,and improve the validity of the near neighbor graph.Firstly,this paper gives a brief description to the related content of cluster analysis,including the research background,significance and the staus of the research.We also systematically introduced the theoretical knowledges of definition,theorem and so on,which are involved in the clustering algorithm.Aiming at the existing problems in graph clustering algorithm,we propose a clustering model:Robust Adaptive Graph-Regularization Clustering Algorithm?RAGR?,that will construct similar data matrix and clustering at the same time based on₂L_1,-norm and graph regularization idea.The algorithm replaces the F-norm with L_2,1-norm,which can not only ensure the sparse data line,but also keep the rotation invariance of the matrix and the robustness to the noise data.We give the optimization algorithm of the model,and use matlab software to verify our algorithm on the synthetic data set and the real dataset.The experimental results show the effectiveness of the proposed RAGR clustering algorithm.