Spatial Data Mining Algorithm: A Proposal with Big Data Characteristics

Gouveia WR, Valencio CR, Zafalon GFD and Colombini AC

Published on: 2024-04-29

Abstract

Data mining in spatial Big Data is more complex compared to conventional data mining because of its intrinsic relations, self-correlation, and the difficulty of extracting patterns from this data. The massive volume of spatial data information generates problems such as exponential growth of data volume, contamination, and computational complexity. In this sense, parallelization techniques are constantly used in order to optimize the processing time of this large amount of data. Thus, to provide some contribution within this context, this paper proposes a new algorithm called CHSMST+MR, based on CHSMST, where distributed and parallel processing techniques are used. In comparison with its literature counterpart, an average reduction of approximately 51.36% in the time of the proposed algorithm was observed. Therefore, it was intended as a scientific contribution, the use of distribution and parallelization to create a solution with superior performance in order to extract behaviors and patterns in spatial data and with Big Data characteristics, given that, since more complex data is being dealt with, it requires more computational processing resources.