Modeling Of a Data Driven Flood Detection and Early Warning System Using Machine Learning Technique

Ndulue, Christopher and Vincent, C. Chijindu and Edward, C. Anoliefo and Eneh, Joy N. (2023) Modeling Of a Data Driven Flood Detection and Early Warning System Using Machine Learning Technique. International Journal of Advances in Engineering and Management (IJAEM), 5 (8). pp. 400-409. ISSN 2395-5252

Text journal paper.pdf Download (477kB)

Abstract

This research presents modeling of a data driven flood detection and early warning system using machine learning technique. The aim is to save lives and properties, through detection of flood, early notification and warning for immediate evacuation. To achieve this, Anam community in Anambra State, Nigeria, which is annually characterized with flood since 1966 till date was considered as the case study. Literatures were reviewed on flood detection systems and lack of reliability was identified as a research gap. To address the gap, the methodology used are hydrological flood modeling, data collection, explorative data analysis approach, data processing with multiple imputation approach, machine learning algorithms (Linear Regression (LR), Random Forest (RF) and Decision Tree (DT)), early warning system based alarm notification. The models were implemented with Python programming language. The results of the machine learning flood detection models were evaluated considering recall, accuracy, receiver operator characteristics curve and compared after tenfold cross validation. The average recall obtained for LR is 99.146%, average accuracy of flood detection is 91.87% and average ROC result which is probability of correct flood detection is 83.23%. Similarly the validation result of the LR detection model for flood early detection reported average recall of 88.42%, average accuracy of flood detection is 92.409% and average ROC result which is probability of correct flood detection is 83.44%. In the same vein, the DT reported average recall of 77.77%, average accuracy of flood detection is 75% and average ROC result which is probability of correct flood detection is 72.22%. The table 5 was used to compare the models to recommend the best which was used for the modeling of the new system

Item Type:	Article
Subjects:	A General Works > AC Collections. Series. Collected works
Divisions:	Faculty of Arts > Faculty of Law > Faculty of Management and Social Sciences > Faculty of Education
Depositing User:	mrs chioma hannah
Date Deposited:	08 Sep 2023 08:41
Last Modified:	08 Sep 2023 08:41
URI:	http://eprints.gouni.edu.ng/id/eprint/4143

Actions (login required)

View Item