Abstract: One of the most catastrophic natural disasters that are extremely difficult to model are floods. Research on improving flood prediction models has helped to lower risks, recommend policy changes, reduce the number of fatalities, and lessen property damage from floods. Over the past two decades, research has shown that machine learning (ML) methods have greatly advanced prediction systems, offering better performance and cost-effective solutions. These methods include logistic reasoning, decision trees, support vector classification, KNN classifiers, and random forest classifiers. The process of modelling the likelihood of a discrete result given an input variable is known as logistic regression. Each internal node of a decision tree, which resembles a flowchart, represents a "test" on an
Flood alerts, flood reduction, and flood avoidance are all possible benefits of using machine learning (ML) models for flood prediction. Due to their cheap computational requirements and predominance of observational data, machine-learning (ML) approaches have grown in popularity as a result. The goal of this study was to develop a machine learning model that can forecast floods in the Nigerian state of Kebbi using historical rainfall data from the previous 33 years (33). This model may then be applied to other high-risk flood-prone states in Nigeria. This paper assessed and compared the accuracy, recall, and receiver operating characteristics (ROC) scores of three machine learning algorithms: decision trees, logistic regression, and support vector classifiers (SVR). Compared to the other two techniques, logistic regression produces higher accuracy outcomes.
|
DOI:
10.17148/IJARCCE.2022.11798
