Proposal of Three Algorithms Improving the DENCLUE Algorithm for Data Clustering

One of the most widely used techniques in Machine Learning is clustering. It is an approach for processing unlabeled data to identify homogeneous groups. This approach requires the construction of efficient models to develop high-performance algorithms. The use of clustering methods allows data to be organized into more manageable categories, thus simplifying the analysis, understanding and interpretation of the information contained in this complex data. This approach has prompted the development of new algorithm alternatives aimed at improving data clustering. In this perspective, various clustering techniques have been put forward, notably the DENCLUE algorithm, which uses the density of the data to detect clusters. It relies on the Hill Climbing algorithm to provide support in the crucial phase of class reconstruction. The aim of this paper is to evaluate the performance of this algorithm using strategies based on the three Hill Climbing variants (Simple Hill Climbing, Steepest Ascent Hill Climbing and Stochastic Hill Climbing) on the different data sets.