Construção de um banco de dados de experimentos de índices espaciais para treinamento de algoritmos de aprendizado de máquina

Abstract

Spatial databases have been widely used in myriad applications of our daily lives, from locationbased services to precision agriculture systems. The growing volume of data generated and consumed by these applications and combined to the existing complexity in handling this type of information makes it necessary to study optimization techniques to query this data in the most efficient way possible. One of the most used techniques is spatial indexing, whose main objective is to reduce the search space and consequently the computational processing. However, there are several indices in the literature and different types of configurations associated with them, making a certain index combined with a certain parametrization obtain better performance than others on a database. In this sense, the objective of this work is to execute experimental tests to collect statistical information from indexed databases and build a dataset with characteristics necessary for a machine learning algorithm that will be able to be trained and predict the best index and its settings for a given spatial database.