An ETL-based system for collecting and processing geospatial data for machine learning analysis

This paper discusses modern tools and approaches to the design of ETL systems for the integration and preprocessing of raster, vector, and tabular geospatial data. These data can subsequently be used for analysis using machine learning methods. The proposed approach is based on a modular pipeline architecture adapted to the characteristics of GIS data formats. It supports integration with machine learning pipelines without requiring costly licenses or substantial modifications of existing platforms.

A specialized ETL pipeline based on the OpenNeurons platform and PostgreSQL/PostGIS database is presented. The system includes configurable modules for describing data sources and adapters implemented in Python using libraries such as geopandas, rasterio, and pyshp. The system also includes a driver for direct interaction with file storage and scheduled data updates. It further provides a web interface for task management, monitoring, and logging. The developed system was tested on real-world geospatial datasets. Firstly, a unified dataset containing 227 geological and geophysical attributes for the eastern sector of the Russian Arctic was created and prepared for machine learning analysis. Secondly, binary classification of 240 morphostructural nodes in the Caucasus region using Gradient Boosting and AutoKeras, an F1-score close to 1.0 was achieved on the test set. Comparison with the KORA-3 algorithm showed a 67% agreement in predictions, indicating the potential applicability of the proposed approach, although further validation on independent datasets is required. Compared to general-purpose ETL solutions, the proposed system offers low deployment costs, high extensibility through custom adapters, and tight integration with machine learning frameworks. The results demonstrate the applicability of the developed ETL pipeline for geospatial data processing tasks.

1. Rakotoarison T., Hajalalaina A.R. Real-time geo-decisional system for risk and disaster management in Madagascar. Reports on Geodesy and Geoinformatics, 2025. Vol. 119. Pp. 7–13. DOI: 10.2478/rgg-2025-0002.

2. Tedikova A.A., Klimochenkov M.D., Melnichenko I., Krasnotsvetov M.A., Us S.S., Kutlyev I.I., Shchekina M.V. The history of origin and development of geoinformatics as a science. Mining Industry Journal (Gornay Promishlennost), 2024. Pp. 90–99. DOI: 10.30686/1609-9192-2024-3-90-99.

3. Rozenberg I., Dulin S. Current Issues Problems of Geoinformatics. Russian Journal of Earth Sciences, 2024. Vol.24. no.1. Pp. 1–11. DOI: 10.2205/2024ES000893.

4. Lisenkov I.A., Solov'ev A.A., Kuznecov V.A., Nikolova YU.I. Obobshchennyj massiv geologo-geofizicheskoj informacii vostochnogo sektora Rossijskoj Arktiki dlya provedeniya analiza metodami mashinnogo obucheniya [A generalized array of geological and geophysical information from the eastern sector of the Russian Arctic for analysis using machine learning methods]. Geologiya i geofizika, 2025. Vol. 66, no. 2. Pp. 232–246. (in Russian) DOI: 10.15372/GiG2024148.

5. Cheruku S., Goel O., Jain S. A Comparative Study of ETL Tools: DataStage vs. Talend. Journal of Quantum Science and Technology, 2024. Vol.1. Pp. 80–90. DOI: 10.36676/jqst.v1.i1.11.

6. Al-Yadum S., Xion T.E., Wei S.G.W., Boursier P. Review on Integrating Geospatial Big Datasets and Open Research Issues. IEEE Access, 2021. Pp. 1–1. DOI: 10.1109/ACCESS.2021.3051084.

7. Astriani W., Trisminingsih R. Extraction, Transformation, and Loading (ETL) Module for Hotspot Spatial Data Warehouse Using Geokettle. Procedia Environmental Sciences, 2016. Vol. 33. Pp. 626–634. DOI: 10.1016/j.proenv.2016.03.117.

8. Borkowski A., To Duc A. Integration of BIM and GIS Data of a Heritage Building Using FME // Civil and Environmental Engineering Reports, 2024. Vol. 34. Pp. 204–215. DOI: 10.59440/ceer/190231.

9. Lisenkov I.A., Soloviev A.A., Kuznetsov V.A., Nikolova Yu.I. Generalized Dataset of Geological and Geophysical Information of the Eastern Sector of the Russian Arctic . ESDB repository, GCRAS, Moscow, 2024. DOI: 10.2205/public-artcic-geospatial-data.

10. Soloviev A.A., Novikova O.V., Gorshkov A.I., Piotrovskaya E.P. Raspoznavanie raspolozheniya potentsial'nykh ochagov sil'nykh zemletryaseniy v Kavkazskom regione s ispol'zovaniem GIS-tekhnologiy [Recognition of the location of potential sources of strong earthquakes in the Caucasus region using GIS technologies]. Doklady Akademii Nauk, 2013. Vol. 450, No. 5. Pp. 599–601 (in Russian). DOI: 10.7868/S0869565213170222.

11. Gvishiani A.D., Dzeboev B.A., Agayan S.M. Intellektual'naya sistema raspoznavaniya FCAZm v opredelenii mest vozmozhnogo vozniknoveniya sil'nykh zemletryaseniy gornogo poyasa And i Kavkaza [Intelligent FCAZm pattern recognition system for identifying possible locations of strong earthquakes in the mountain belts of the Andes and the Caucasus]. Fizika Zemli, 2016. No.4. Pp.3–23 (in Russian). DOI: 10.7868/S0002333716040013.

12. Soloviev A.A., Gorshkov A.I., Soloviev A.A. Primenenie dannykh po litosfernym magnitnym anomaliyam v zadache raspoznavaniya mest vozmozhnogo vozniknoveniya zemletryaseniy [Application of lithospheric magnetic anomaly data in the problem of identifying possible locations of earthquakes]. Fizika Zemli, 2016. No. 6. Pp. 21–27 (in Russian). DOI: 10.7868/S0002333716050148.

13. Gelfand I.M., Guberman Sh.A., Keylis-Borok V.I., Knopoff L., Press F., Rantsman E.Ya., Rotvain I.M., Sadovsky A.M. Usloviya vozniknoveniya sil'nykh zemletryaseniy (Kaliforniya i nekotorye drugie regiony). Issledovanie seysmichnosti i modeley Zemli [Conditions for the occurrence of strong earthquakes (California and some other regions). Study of seismicity and Earth models]. Vychislitel'naya seysmologiya, 1976. Iss.9. Pp. 3–91 (in Russian).