An Evolutionary MultiLayer Perceptron Algorithm for Real Time River Flood Prediction

An Evolutionary MultiLayer Perceptron Algorithm for Real Time River Flood Prediction

Table of Contents


Severe flash flood events give very little opportunity for issuing warnings. In this paper, we approach the automated and real-time prediction of river flooding by proposing and evaluating different variations of the conventional Multilayer Perceptron (MLP) machine learning algorithm. Our first approach follows a trial and error attempt to optimize MLP architecture. The second and third approaches are based on the application of nature-inspired evolutionary techniques, namely the Genetic Algorithm (MLP-GA) and the Bat Algorithm (MLP-BA) respectively. The MLP-GA generates an improved MLP configuration and MLP-BA enhances the training method. Our fourth, novel approach (MLP-BA-GA) is based on the application of GA to further optimize both the BA and MLP architecture. When compared with previous work, experiments show improvement in the accuracy of river flood prediction, with significant results for the MLP-BA-GA.

  • Author Keywords

    • Artificial Neural Network,
    • Bat Algorithm,
    • Flood Prediction,
    • Genetic Algorithm,
    • Machine Learning Optimization,
    • Metaheuristic Model,
    • Evolutionary Computing
  • IEEE Keywords

    • Training,
    • Genetic algorithms,
    • Floods,
    • Machine learning algorithms,
    • Computational modeling,
    • Classification algorithms,
    • Artificial neural networks


Global warming is responsible for changes in precipitation patterns causing severe weather conditions such as torrential rainfall and flood. Several regions around the globe deal with the sudden rise of water level causing damage to infrastructure and the loss of human life. This situation, known as a flash flood, is even more important for less developed and developing countries, causing drastic impacts on their economy. On a worldwide scale from 1980 to 2009, it is estimated that flood was responsible for more than 500,000 loss of life and affected more than 2.8 billion people [1]. Among different approaches used to monitor floods, we focus on the use of Wireless Sensor Networks (WSN). Different parameters, such as water level, temperature, radiation can be monitored in near real-time. This large set of collected data is analyzed to forecast the possible occurrence of a flood and such predictions can be lifesaving. It is therefore important to automate this process in an attempt to minimize human intervention and avoid unnecessary delays. Several Machine Learning (ML) algorithms have been used towards this end [2, 3], as they can analyze the collected data, which have been pre-processed, and generate learning models to make forecasts autonomously. It has been demonstrated by Furquim et al. [2] that traditional Machine Learning techniques, such as Bays Net, MLP, J48, Random Forest, Random Tree, and BF Tree provide different levels of prediction accuracy. Some of which are quite appropriate for flood prediction, but more recent studies, in different other domains, using hybrid techniques provide even better results [7, 8, 9] than standalone approaches. We believe that further optimization of Machine Learning techniques, especially Artificial Neural Networks (ANNs) is possible for better prediction of flash flood nowcasting with the combination of the Genetic Algorithm (GA) and nature-inspired techniques. An ANN emulates the human brain consisting of different layers such as the input layer, processing (hidden) layer(s), and the output layer. The layers are made up of nodes (neurons) connected by weighted edges (synapses). Feed-forward propagation is the process by which data are fed in from input nodes to the output nodes. Backpropagation is the learning process by which the output error of the network is propagated backward through the network and weights are adjusted. GAs are typically used to optimize ANNs and are inspired by the natural selection process, where the fittest individuals are selected for producing offspring for the next generation. The network parameters for the offspring are inherited from the fittest parent network in the previous generation. Swarm based nature-inspired algorithms such as the Bat Algorithm (BA) is an emerging optimization approach that can further enhance the prediction accuracy of Machine Learning techniques. BA is based on the echolocation of micro-bats which allows them to collectively recognize tiny prey among the various existing living and non-living objects, in night darkness [10]. It, therefore, combines the capabilities of both population-based and local search techniques. With its strong local searchability, the algorithm can converge to a global optimum, enhancing machine training aspects like feature selection, hidden layers, and a number of nodes per layer [11, 12]. In this research work, we propose to analyze and evaluate the following Multi-Layer Perceptron (MLP) metaheuristics hybrid models: Optimized MLP, MLP-GA, MLP-BA, and MLP-BA-GA.


The results demonstrate that our hybrid Metaheuristic Machine Learning algorithm (MLP-BA-GA) increases the accuracy of prediction results. It is of main importance for time-critical tasks, especially when human lives are at stake due to natural calamities like river floods. The MLP-BA-GA is recommended in cases where training can be performed offline, as this process imposes an intensive use of computer resources and is time-consuming. Further improvement can be achieved by implementing a feature selection algorithm to filter out unnecessary features. It has the potential to further improve accuracy as well as reduce the computation complexity which in turn lowers the computation time. Also, an investigation of different sets of window sizes and data attributes can also contribute to improving the nowcasting accuracy. From a deployment perspective, the proposed approach can take the form of a web service with a REST interface. As such, mobile and web applications can be developed and receive updates with regards to the possibilities of river overflowing. It can benefit communities that live close to rivers as well as the necessary authorities responsible for dealing with natural calamities.

About KSRA

The Kavian Scientific Research Association (KSRA) is a non-profit research organization to provide research / educational services in December 2013. The members of the community had formed a virtual group on the Viber social network. The core of the Kavian Scientific Association was formed with these members as founders. These individuals, led by Professor Siavosh Kaviani, decided to launch a scientific / research association with an emphasis on education.

KSRA research association, as a non-profit research firm, is committed to providing research services in the field of knowledge. The main beneficiaries of this association are public or private knowledge-based companies, students, researchers, researchers, professors, universities, and industrial and semi-industrial centers around the world.

Our main services Based on Education for all Spectrum people in the world. We want to make an integration between researches and educations. We believe education is the main right of Human beings. So our services should be concentrated on inclusive education.

The KSRA team partners with local under-served communities around the world to improve the access to and quality of knowledge based on education, amplify and augment learning programs where they exist, and create new opportunities for e-learning where traditional education systems are lacking or non-existent.

FULL Paper PDF file:

An Evolutionary MultiLayer Perceptron Algorithm for Real-Time River Flood Prediction



G. Suddul, K. Dookhitram, G. Bekaroo, and N. Shankar,




An Evolutionary MultiLayer Perceptron Algorithm for Real-Time River Flood Prediction

Publish in

2020 Zooming Innovation in Consumer Technologies Conference (ZINC), Novi Sad, Serbia, 2020, pp. 109-112, 



PDF reference and original file: Click here

+ posts

Somayeh Nosrati was born in 1982 in Tehran. She holds a Master's degree in artificial intelligence from Khatam University of Tehran.

Website | + posts

Professor Siavosh Kaviani was born in 1961 in Tehran. He had a professorship. He holds a Ph.D. in Software Engineering from the QL University of Software Development Methodology and an honorary Ph.D. from the University of Chelsea.

Website | + posts

Nasim Gazerani was born in 1983 in Arak. She holds a Master's degree in Software Engineering from UM University of Malaysia.