Аннотация:During the last ten years, machine learning approach for data analysis has been a topic of
growing interest for the scientific community. Not surprisingly so, since the last decade it has
brought about a true revolution in the world of technology with new and extremely advanced
approaches to computer modeling. As far as engineering depends on data processing it is very
important to provide it with modern and reliable analytical systems.
Nowadays, implementation of geo-information systems within the field of Earth Sciences and
affiliated parts of industry is highly relevant. Particularly for engineering surveys it gives a
tremendous potential for data analysis. In addition, rapidly growing open source community
of data science provides us a chance to implement new and flexible software for data mining.
The present study makes an attempt to combine basic principles from GIS modeling and
machine learning techniques in order to develop sustainable prediction systems for hazardous
events. The models are based on the results of engineering surveys and remote scanning
within the Taman peninsular. Particularly, they describe distribution of the most hazardous
processes for the buildings construction: collapsibility and landslides. The final model is
aimed to determine suitable sites for civil development based on geological safety and
resource economy perspective. We explore nonlinear connections between different variables
using the most up-to-date methods of computational mathematics in order to realize full
potential of our data.