An Improved Semi Supervised Clustering Of Given Density And Its Application In Lithology Identification

With the rapid development of information technology,the scale of data has exploded.Finding valuable information from intricate data is of great practical significance.As an important method in the field of machine learning,clustering algorithms are widely used in data analysis and mining.DBSCAN algorithm is a typical clustering algorithm.Data density can be used as a measure to identify the shape and noise points of any shape in the data set.However,the algorithm uses fixed parameter clustering in the clustering process,and the clustering effect on non-uniform density samples will be greatly reduced.This paper takes DBSCAN algorithm as the research object,and studies the problems of sensitive to clustering parameters and unsatisfactory results when dealing with non-uniform density data sets,and proposes an improved method V-DBSCAN algorithm.V-DBSCAN can be better suited for non-uniform density clustering.The main idea of this algorithm is to first find high-density clusters,then change the parameter values to continue clustering,and discover lower-density clusters.Constantly changing the value of the parameter,clusters of different densities will be distinguished.In the continuous clustering process is also accompanied by clustering.At this time,a semi-supervised learning method is added,and the generated clusters are merged with the tag information or the constraint information of the known point to improve the speed and accuracy of the merging process.At the same time,an R-tree index was established to improve the operating efficiency of the program.Because of the complexity of the geological reservoir data,the traditional method of logging lithology identification is not ideal,so the improved method is applied to the lithology identification,which is used as the practical application of the improved algorithm.The experiment used the V-DBSCAN algorithm for logging lithology identification.Three different distances were compared with K-means,KNN,and NBC algorithms.The differences in the accuracy of these four algorithms in lithology recognition were analyzed.The experimental results show that the accuracy of the algorithm used in lithology identification of logging data is higher than other algorithms,which proves that V-DBSCAN can be well applied to logging lithology identification.