A Novel Approach for Classifying the Management Priorities of Flooding Events Using Clustering Algorithms and Geospatial Analysis

Due to the frequent occurrences and damages of flooding disasters, there has been an increasing interest in developing proper methods to help mitigate their consequences. Although previous research was directed to help in managing disasters, more data-driven methods are still needed. To this end, this paper developed a novel approach to enhance the decision-making process related to prioritizing the flooding mitigation, management, control, and/or recovery plans. First, data was collected for multiple flooding events and was cleansed to reach a total of 7,152 observations. Second, exploratory data analysis was conducted to examine and uncover trends and relationships in the data. Third, unsupervised machine learning was used to categorize the management priority of the disaster events using clustering analysis. Fourth, geospatial analysis was conducted on both the flood event level and on the county level. The results showed that flood disaster events could be categorized—based on their duration and frequency—into two management priority levels: low priority and high priority. The conducted research in this paper contributes to the body of knowledge by equipping agencies and disaster decision-makers with a decision-support system to prioritize short to long-term risk reduction and management interventions to better address flood disaster events.