2026-05-20T18:28:25.979Z http://ezaposleni.singidunum.ac.rs/rest/sciNaucniRezultati/oai

ezaposleni.singidunum.ac.rs/rest/sciNaucniRezultati/oai:3:9807 2024-12-03T20:10:47Z 3

Flood Prediction Based on Recurrent Neural Network Time Series Classification Boosted by Modified Metaheuristic Optimization, Chapter in LNNS Lecture Notes in Networks and Systems: ADCIS 2023: Advances in Data-Driven Computing and Intelligent Systems, Springer, volume 893 2024 http://ezaposleni.singidunum.ac.rs/rest/sciNaucniRezultati/oai/record/3/9807 https://link.springer.com/chapter/10.1007/978-981-99-9518-9_21 I. Markovic J. Krzanovic L. Jovanovic A. Toskovic N. Bacanin A. Petrovic M. Zivkovic Floods are a type of natural calamity that can result in substantial harm to the surroundings, built structures, and, regrettably, even the loss of human lives. However, with timely warning actions, proactive measures can be taken to mitigate damage and prepare adequately, thereby preventing major losses. This work delves into the exploration of artificial intelligence algorithms to enhance flood prediction capabilities by leveraging historical time series data. To achieve this, a recurrent neural network model has been engaged and applied to actual data for analysis and evaluation. In addition, success of the neural network heavily relies on the selection of proper hyperparameters. Hence, a modified version of a well-known optimization metaheuristic is introduced to optimize the tuning process, aiming to leverage the model’s accuracy and reliability. To rate the effectiveness of the introduced method, it is compared against several established optimization algorithms, considering factors such as computational efficiency and prediction accuracy.This comparative examination sheds light on the strengths and weaknesses of each method, aiding in the identification of the most effective approach for flood prediction and mitigation. The suggested method attained excellent accuracy of 72.8%, with superior values of precision, recall and F1-score measurements especially for the flood class. bookPart Springer, Singapore 289 303 10.1007/978-981-99-9518-9_21 LNNS Lecture Notes in Networks and Systems: ADCIS 2023: Advances in Data-Driven Computing and Intelligent Systems, volume 893