A novel approach for high-velocity big geo-data handling using iterative and feature learning algorithms

Abstract

Geospatial data were exclusively generated by official agencies. However, following the technological revolution in data collection and production, various sources have emerged for the massive production of geospatial data, resulting the phenomenon of big geo-data. Therefore, dealing with large amounts of these data sets, results in a high velocity as they change very quickly, is a challenging task. Hence, analysis become more complex and computation become prohibitively expensive. As a result, spatial computing technologies become limited in front of these complex data and operations. Accordingly, we aimed to refine complexity with simplicity by replacing traditional geospatial models with referring to the simplest intelligent and minimum resource requirement algorithms that can be applied against these constraints, while ensuring the criteria of performance and scalability. In this work, we focus on the high-velocity of this big geo-data through the use of an iterative approach applied to a feature learning algorithms to decrease the memory consumption and the time complexity of traditional machine learning algorithms. According to our knowledge, although they were widely applied in the 19th century as a solution to overcome the problems of limitation of memory and computing resources. Iterative methods were still not used for the big geo-data analytics and generally for the big data domain. Thus, this approach could be beneficial especially for real time applications such as the anomaly monitoring and detection.