METHODOLOGY FOR DIGITAL MAPPING OF POSSIBLE TERRITORY FLOODING USING SEMI-SUPERVISED MACHINE LEARNING

Abstract

This paper presents a comprehensive methodology for digital mapping of possible territory flooding using semi-supervised machine learning techniques within the QGIS geographic information system. The proposed methodology encompasses several critical stages: data collection, preliminary processing, and the development of predictive models, all aimed at accurately identifying and mapping areas vulnerable to flooding. This methodology uses semi-supervised learning, which effectively leverages marked and unmarked data to enhance the model's predictive capabilities. By utilizing both types of data, the approach significantly improves the quality of forecasting and the accuracy of mapping the occurrence of possible territory flooding.

The methodology ensures that valuable information from various data sources – such as satellite imagery, meteorological data, and GIS – is seamlessly integrated to provide a holistic view of flood risk. A notable feature of the methodology is its adaptability to different regions and contexts, allowing for tailored risk assessments that account for local geographic and climatic conditions. This adaptability is particularly essential in the context of climate change and growing urbanization, which affect the dynamics and intensity of floods. The practical implications of this methodology are substantial and impactful.

It facilitates the timely identification of flood risk zones, which helps mitigate potential negative impacts on communities and infrastructure by providing accurate and timely data for effective planning and decision-making. Additionally, the integration of semi-supervised learning techniques with QGIS enhances the usability and accessibility of the methodology for a wide range of users, including researchers, policymakers, and disaster management professionals.