Time Series Analysis of Geospatial Data

This graduation paper is devoted to the research and application of methods of data analysis in space and time. The methods have been applied to an actual spatiotemporal dataset (bird observations). This work is relevant due to the wide spread of geoinformation systems and methods’ universality. Besides, there is a lack of studies focused on spatiotemporal analysis compared to works devoted to spatial and temporal analysis individually. Moreover, there are few ornithological studies that could contribute to ecosystem preservation in Russia. In the course of work scientific literature has been studied; access to the eBird database has been obtained; a visualization tool has been implemented (with an additional feature to visualize clusters, computed by the space-time permutation model); we conducted a comparison of linear regression and neural network models, predicting the date of birds’ arrival. It has been confirmed that neural networks perform better dealing with the problem.