COMPARATIVE ANALYSIS OF K-MEANS, K-MEANS++ AND MINI BATCH K-MEANS ALGORITHMS IN PYTHON ENVIRONMENT

This article discusses the application of k-means algorithm and its modifications to datasets of different dimensions in the Python environment. At the same time, the current state, opportunities, shortcomings, problems of the traditional k-means clustering algorithm and its modifications are studied and suggestions for their solution are given. The k-means ++ algorithm eliminates the disadvantage of the traditional k-means method’s random selection of starting centers. Using the mini-batch k-means algorithm, big data is analyzed by dividing it into packets, which accelerates the process of analyzing large and complex data. A hybrid PCA and elbow method are proposed to reduce the dimensionality during data clustering and to find the optimal number of clusters. To evaluate the effectiveness of this approach, algorithms are tested on several sets of data of different sizes. The silhouette and Davis-Boldin indices are used to evaluate the efficiency of the algorithms. The results of the experiment show that the proposed approach is more efficient when clustering big data. The proposed hybrid PCA and elbow method create new opportunities for solving problems that require large computing resources in the process of analysing large, multidimensional data (pp.119-128).