Situations that generate a continuous data stream, such as TCP / IP traffic, e-commerce, and industrial monitoring, can make the usability of algorithms that have machine learning completely offline unviable. That is due to the need for data storage, the infinite growth of data generation, and limited memory restrictions. With that, the algorithms that have the learning totally or partially on-line appeared. Among them, there are the evolving algorithms, which have been of interest because they can develop and update in unknown environments and detect concepts drift and evolution in the input data over time. Because of these algorithms' broad applicability in real problems, this work proposes a new evolving algorithm named Macro SOStream. This algorithm has on-line learning and is based on self-organizing density for data stream clustering. The Macro SOStream is based on the SOStream algorithm, but we incorporated macroclusters composed of the microclusters. While microclusters have spherical shapes, macroclusters can have arbitrary shapes. Besides, the Macro SOStream's performance was compared to SOStream and DenStream algorithms' performance using the datasets and the ARI performance metric to validate our proposal.