IOP Conference Series: Earth and Environmental Science

PAPER • OPEN ACCESS

Flood Forecasting using Committee Machine with Intelligent Systems: a

Framework for Advanced Machine Learning Approach
To cite this article: Amrul Faruq et al 2020 IOP Conf. Ser.: Earth Environ. Sci. 479 012039

View the article online for updates and enhancements.

This content was downloaded from IP address 183.88.58.20 on 04/11/2020 at 10:32

The 7th AUN/SEED-Net Regional Conference on Natural Disaster (RCND 2019) IOP Publishing
IOP Conf. Series: Earth and Environmental Science 479 (2020) 012039 doi:10.1088/1755-1315/479/1/012039

Flood Forecasting using Committee Machine with Intelligent

Systems: a Framework for Advanced Machine Learning
Approach

Amrul Faruq*1, 2, Shahrum Shah Abdullah*2, Aminaton Marto3 , Che Munira Che Razali2,
and Shamsul Faisal Mohd Hussein2

1
Department of Electrical Engineering, Faculty of Engineering, Universitas Muhammadiyah
Malang, Indonesia
2
Department of Electrical and Electronics Systems, Malaysia-Japan International Institute of
Technology, Universiti Teknologi Malaysia, Malaysia
3
Department of Environmental Engineering and Green Technology, Malaysia-Japan
International Institute of Technology, Universiti Teknologi Malaysia, Malaysia

*Corresponding author e-mail: [email protected], [email protected]

Abstract. Among many natural hazards, flood disasters are the most incisive, causing
tremendous casualties, in-depth injury to human life, property losses and agriculture, therefore
affected the socioeconomic system of the area. Contributing to disaster risk reduction and the
property damage associated with floods, the research on the advancement of flood modelling
and forecasting is increasingly essential. Flood forecasting technique is one of the most
significant current discussion in hydrological-engineering area, in which a highly complex
system and difficult to model. The past decade has been seen the rapid development of machine
learning techniques contributed extremely within the advancement of prediction systems
providing better performance and efficient solutions. This paper proposes a framework design
of flood forecasting model utilizing committee machine learning methods. Previously published
works employing committee machine techniques in the analysis of the robustness of the model,
effectiveness, and accuracy are particularly investigated on the used in various subjects. It is
found that artificial neural networks, hybridizations, and model optimization are reported as the
most effective ways for the improved development of machine learning methods. The proposed
framework employs four representative intelligent systems as individual members, including
radial basis neural networks, adaptive-neuro fuzzy, support vector machine and deep learning
networks to construct a committee machine. As a conclusion, this committee machine with
intelligent systems appears to be capable of enhancing the designing of flood forecasting model
for disaster risk reduction.

1. Introduction
Flood disasters keep on happening in numerous nations around the globe due to the dynamic climate
change condition. Among the natural hazards, flood disasters are the most destructive. Huge flood
causing tremendous casualties, extensive damage to human life, property losses, agriculture and the
socioeconomic system. In order to reduce the impact of this disaster, the governments, therefore, are
under pressure to develop and provide an accurate and robust flood forecasting for disaster risk
management [1]. Flood forecasting models are important in hazard assessment and extreme event
management. The research on the advancement of flood forecasting is increasing since it contributes to
disaster risk reduction, which is a difficult task, challenging and highly complex to model [2]. According
Content from this work may be used under the terms of the Creative Commons Attribution 3.0 licence. Any further distribution
of this work must maintain attribution to the author(s) and the title of the work, journal citation and DOI.
Published under licence by IOP Publishing Ltd 1
The 7th AUN/SEED-Net Regional Conference on Natural Disaster (RCND 2019) IOP Publishing
IOP Conf. Series: Earth and Environmental Science 479 (2020) 012039 doi:10.1088/1755-1315/479/1/012039

to Sendai Frameworks 2015-2030, disaster risk reduction (DRR) is given by priority number three and
four, which are ‘investing in disaster risk reduction for resilience’ and ‘enhancing disaster risk
preparedness for effective response’, respectively [3]. In connection with this viewpoints, hence flood
modelling and forecasting is crucial for disaster risk management. In many regions of the world, flood
forecasting is one among the few feasible options to manage flood disasters.
To date, a number of flood forecasting models are mainly data-specific and involve simplified
various input assumption [4]. Thus to mimic the complex mathematical expression of physical processes
and river behaviour, such models benefit from specific techniques, e.g., empirical black-box models,
stochastic and hybrids [5]. These physically and statistically based models boost the usage of advanced
data-driven methods, e.g., Machine Learning (ML) and Deep Learning (DL) technique. Data-driven
forecasting methods using ML are promising tools as they are less time consuming to develop with
minimal inputs. ML technique is one of the most significant current discussion in Artificial Intelligence
(AI) fields. Among them, the most well-known works of flood forecasting modelling include artificial
neural networks (ANNs) [6], support vector machine (SVM) [7] adaptive neuro-fuzzy inference system
(ANFIS) [8], were effectively employed for both short-term and long-term flood forecasting. As a new
method in ANN models, deep learning is a major subject of interest within the field of AI methods. Deep
learning is being studied in many types of problems such as image processing, speech recognition, and
natural language processing. In the subject of forecasting, recent studies have been reported the
successful use of deep learning in various fields [9], [10], [11] respectively for power load and
probability density forecasting, traffic flow forecasting and rainfall forecasting. As it developed that
deep learning proven reported better result than traditional ANN model [12].
Previous methods are indicative of all individual models being capable of forecasting the floods.
Different AI models provide a similar acceptable efficiency but with different characteristic strengths
and weaknesses. So that, exploiting the synergy among better performing models is an attractive
proposition if the positive aspects of different modelling techniques can be combined. One such
technique is Intelligence Committee Machine (ICM) or sometimes called Committee Machine with
Intelligent System (CMIS) models that were explored in various disciplines; river flow forecasting, gas
reservoirs and rock permeability predictions [13], [14], [15], respectively. It introduces an AI-based
multi-model interface to exploit their synergy. This uses outputs from different AI models and
determines the interface to reach the overall decision on identifying better performing AI models. Using
ensemble committee-based data intelligent approach, researchers have successfully employed for
generating soil moisture forecasts [16]. The CMIS combines AI model results by simple ensemble
averaging [17] or by weighted averaging, which is adopted via optimization methods such as Genetic
Algorithm (GA) [18]. Gholami et., al. [19] compared GA and simple ensemble averaging method as
combiners and concluded that the GA is more efficient. Notably, the term committee is understood to
refer generally to the synergic combination of a few models and machine to be another word for
artificial. The advantage of the CMIS is a capability for a nonlinear combination of AI models under
supervision leading to improvements in the performance of CMIS over individual AI models.
The forecasting of flood lead-time and location occurrence is fundamentally sophisticated due to the
dynamic nature of the monsoon phenomenon. Although extensive studies have been carried out on
hydrological-flood forecasting models, there have been very few identified approach that can generally
be applied particularly in AI which was applicable for all types of modelling (e.g., forecasting,
optimization, classification, etc.). Previously published studies are limited to one flood forecasting
model employed in one reservoir, and there was no single AI technique that was suitable for all specific
problems in general [10]. However, the nature of the presented models remains unclear and flood peak
needs to be forecasted more accurately. Along with this growth of forecasting techniques in hydrological
data, all of these applied models still have a notable degree of shortcoming about their generalization
and implementation as an expert system. Therefore, the design of flood modelling remains a passionate
challenge that continues to be undertaken by researchers or scientists.
Investigating multi-model is a continuing concern within the field of advanced machine learning
methods. It has been reported that model integration of intelligent systems and the concept of committee
machine can improve and optimized performance than the individual model. Although studies have
recognized the concept of committee networks, as investigated recently by [20] and [21] the use of

2
The 7th AUN/SEED-Net Regional Conference on Natural Disaster (RCND 2019) IOP Publishing
IOP Conf. Series: Earth and Environmental Science 479 (2020) 012039 doi:10.1088/1755-1315/479/1/012039

CMIS based machine learning models are largely unnoticed in engineering-hydrological science
especially for flood forecasting. A further study as suggested by [22] then to focus on the use of advance-
soft computing methods. The CMIS technique in order to obtain a better result of flood forecasting,
therefore, is proposed in this study. A CMIS has a parallel framework that produces a final output by
combining the results of individual models. These are consist of famous models that widely employed
in ML methods which include ANN, a hybrid neural network and fuzzy system, and support vector
machine. To eliminate the limitation of the usage of a single ANN model, an extending model into deep
learning will also be examined as an individual expert member in that particular CMIS.

2. Application of Committee Machine in Various Study

The study of hydrological data processing, which includes flood hydrograph forecasting, has grown
significantly since the early 1990s where physical-based models were long used to predict hydrological
events such as storm, rainfall-runoff, streamflow forecasting and including floods [23]. Although
physical-based models showed great capabilities for forecasting a diverse range of flooding scenarios,
they often require various types of hydro-geomorphological monitoring datasets, requiring intensive
computation, which prohibits short-term forecasting [24]. Due to these constraints, a large number of
forecasting models that enhanced efficiency using available historical data are developed by the
researchers, which are more robust and versatile [25].
Recently, the use of advanced data-driven methods, including AI models has been attracting
considerable interest in flood forecasting problems. As a result, the practical research on flood
forecasting based on AI models has significantly better performance result compared to the traditional
approach. Jabbari and Bae [26] enhanced the accuracy of real-time flood forecasting using artificial
neural network (ANN) models. Such ML algorithms like SVM [27], and hybrid ANFIS [28] were
reported to be effective techniques for flood forecasting. Furthermore, a recent investigation by Taifur,
et., al. [29] which employed a number of ML forecasting models showed good forecasting result of a
flood using substantially less data, such as easily measurable flow stage. In order to eliminate the
limitation of using single models, an extended model such as deep learning has been derived [30]. More
recently, Caihong Hu, et., al. [30] introduced a state-of-the-art of Long Short-Term Memory (LSTM) as
deep learning method in hydrological time-series data forecasting. Based on the simulation performance,
LSTM models outperformed existing ANN model and found to be more stable.
In order to improve the performance accuracy and to achieve better dataset management, a multiple
of ML modelling options were introduced for hydrological data in the last few years. Wang Bin, et, al
[31] have demonstrated multi-model ensemble (MME) schemes to forecast historical monthly rainfall
and temperature with machine learning methods, the MMEs obtained a better result than any individual
model and can be more efficient and useful having improved performance accuracy. On the other hand,
using the term combination of ML models, Moghadam, et., al. [32] implemented the proposed approach
for flood susceptibility mapping. Although many studies have recognized the use of machine learning
as part of artificial intelligence algorithms, the previous published study has yet to explicitly investigate
the effectiveness of committee machine with intelligent system addressing for floods modelling and
forecasting. The literature related to numerous studies on improving the performance of the model by
using CMIS is summarized in Table 1.

Table 1. The use of CMIS for forecasting application in various fields

Authors/year Application Committee Models Findings
Monomoy Flow forecasting in the Ensemble Average, Used the observed rainfall
Goswami; Kieran absence of quantitative Weighted Average, together with observed river
M. precipitation and Neural Network flow is seen to considerably
O’Connor/2007 improve the performance of
[14] the flow forecasting model.
(a) Ensemble autoregressive
(AR) best in one-two days
forecast.

3
The 7th AUN/SEED-Net Regional Conference on Natural Disaster (RCND 2019) IOP Publishing
IOP Conf. Series: Earth and Environmental Science 479 (2020) 012039 doi:10.1088/1755-1315/479/1/012039

(b) ANN is best in more than

2 days ahead.

Afshin Tatar/2014 Prognosticating Individual MLP; RBF Results reveal the

[33] residual gas saturation and Least Square robustness of CMIS for
in water drive gas SVM; modelling the residual gas
reservoirs CMIS tuned by GA saturation.

Mohammad- Prediction of Stock Ensemble Average; GA and PSO respectively are

Taghi Faghihi- Market Weighted Average; applied in order to optimize
Nezhad; GA and PSO. the direction of the next price
Behrouz Minaei- CMIS tuned by GA movement and create a new
Bidgoli/2018 [34] and PSO training data set. CMIS
shows better result than
individual models.
Asaad Y. River flow forecasting Linear function and The committee with WAM
Shamseldin; MLP ANN and AR shows better result.
Kieran M.
O'connor./2003
[35]
Amir Dashti et Prediction of solubility MLP-ANN; CMIS shows better with
al./2018 [36] of gases within H2- ANFIS; (R2) of 0.9999, 0.9987,
selective GA-ANFIS; 0.9998, 0.9995, and 0.9997
nanocomposite GP Genetic for CMIS, GP, GA-ANFIS,
membranes Programming. ANFIS and ANN
CMIS tuned by GA. models respectively.

Noradin Multi-block engine ANN; The improved fusion

Ghadimi./2018 load and price forecast RBF-NN; algorithm outperforms and
[37] in smart grid SVM . accurate result compared
CMIS tuned by with other forecasting
feature selection and strategy includes ANN,
tuned by chaotic SVM and RBFNN.
shark smell
optimization (C-SSO)
Ali Kadkhodaie- Estimation of Total Fuzzy; CMIS shows better result
Ilkhchi1/2009 Organic Carbon AN-FIS; than individual model
[18] Content from ANN. employed.
Petrophysical CMIS tuned by GA
Parisa Rock permeability MLP; CMIS shows better result
Bagheripour/2014 prediction RBF and Generalised comparing with individual
[15] Regression NN. NN models.
CMIS tuned by GA

Ramendra Prasad, Generating soil M5 model tree; ANN-CMIS shows better

et.,al./2018 [16] moisture forecasts Random Forest; result and performance.
Extreme Learning
Machine.
CMIS tuned by ANN

4
The 7th AUN/SEED-Net Regional Conference on Natural Disaster (RCND 2019) IOP Publishing
IOP Conf. Series: Earth and Environmental Science 479 (2020) 012039 doi:10.1088/1755-1315/479/1/012039

Amin Gholami., Estimation of wax SVR and ANN ; CMIS shows better than
et.,al./2017 [19] deposition CMIS tuned by GA individual models.

M Azmi et Hydrological Multi Linear CMIS tuned by KNN shows

al./2010 [38] Forecasting Regression; better than simple averaging.
MLP-ANN;
KNN ;
CMIS tuned by
ensemble average and
KNN
Robert J. River flow forecasting ANFIS, Fuzzy and CMIS data fusion with NN
Abrahart and ARMA models. shows better performance
Linda See./2002 Each of Rivers than individual model
[13] models tuned by performing alone.
Simple Mean for each
set of forecast;
Single Best
performing;
Neural Network and
Fuzzy Logic
algorithm.

The literature has emphasised the important finding of the effectiveness committee machine models
for prediction application in various discipline of studies. Afshin Tatar et, al. [33] presented the
prediction of residual gas saturation utilizing committee machine intelligent technique in water drive
gas reservoirs using petrophysical data. Three intelligent systems namely radial basis function (RBF),
neural network multilayer perceptron (MLP), and least square support vector machine (LSVM) were
employed. Whereas Nezhad and Bidgoli are more concerned with stock market predictions [34]. To
optimise the combination of the mentioned experts, genetic algorithm (GA) and particle swarm
optimization (PSO) were chosen as weighted averaging technique for its flexibility and well
performance. Result obtained from the developed intelligent approaches more robust and had more
desirable performance. Furthermore, to enhance the precision of ultimate rock permeability prediction,
Parisa [15] was constructed a committee neural network model. The values of rock permeability derived
from the MLP, RBF and generalized regression neural network (GNN) models. While Gholami et, al.
[19] developed a combination of intelligent models through committee machine for the quantitative
estimation of wax deposition. In this paper, committee machine was constructed for combining the
results of the support vector regression (SVR) and ANN models.
In the field of hydrological area, Monomoy and Kieran [14] used a multi-model approach for real-
time flow forecasting in the absence of quantitative precipitation forecasts. The outputs of the models
in this scenario are combined using three techniques of the combination includes simple ensemble
averaging method (SAM), the weighted averaging method (WAM) and the neural network method
(NNM). Azmi et, al. [38] have presented a comparative assessment of five different methods multi-
model data fusion in streamflow and flood peak discharged hydrological forecasting. Data fusion by K-
nearest neighbour (KNN) algorithm was outperformed conventional methods. While Asaad [35]
previously investigated the efficacy of using a combined simulation-mode for real-time river flow
forecasting model. The objective of the approach is to pool the strengths and de-emphasise the
perceptible weakness of the individual models in order to produce ‘consensus’ lead-time flow
forecasting. All of three methods of model output combination produce very similar efficiency values
which are generally better than the efficiencies of the individual models used in combination.
Figure 1 shown numerous schematic diagrams of committee machine developed by the previous
researchers in the various field of problems; prediction of fluoride concentration [39], estimation of total
organic carbon [18], prognosticating residual gas saturation [33], prediction of solubility of gases [36].

5
The 7th AUN/SEED-Net Regional Conference on Natural Disaster (RCND 2019) IOP Publishing
IOP Conf. Series: Earth and Environmental Science 479 (2020) 012039 doi:10.1088/1755-1315/479/1/012039

(a) prediction of fluoride concentration [39] (b) estimation of total organic carbon [18]

(c) prognosticating residual gas saturation [33] (d) prediction of solubility of gases [36]

Figure 1. Schematic diagram of CMIS design used in various disciplines.

3. Proposed CMIS Model Frameworks

There are some reasons for distributing a learning task among a number of individual networks. The
main reason is due to improving the generalization ability because the generalization of individual is not
unique. The combination of some ANNs when they do the same task is called as the ensemble of neural
networks or committee of neural networks. Haykin [40] namely the combination of experts constitute a
committee machine. In this terminology, the author utilizes the combination of ANN models to construct
the committee machine. On the other hand, when the networks are different it is called a committee of
machine, which are ensemble frameworks of single individual machine learning models [41]. Basically
it fuses a knowledge acquired by experts to arrive at an overall decision that supposedly superior to the
attainable by any one of them acting alone.
The proposed methodology comprises of three major steps. At the first stage, the flood water level
will be forecasted from the individual expert as intelligent systems (this study, e.g. RBFNN, ANFIS,
SVM and DCNN). Then a committee machine with this mentioned intelligent system is constructed to
get better generalization functions based on machine learning approach. After the construction of
individual intelligent models, it is necessary to find a suitable method to combine individual results. The
last phase of the design CMIS is the combination of the individual outputs. In this study, ensemble
method based on CMIS design includes ensemble averaging and the weighted averaging algorithm will
be addressed. A schematic diagram of proposed CMIS can be illustrated in Figure 2.

6
The 7th AUN/SEED-Net Regional Conference on Natural Disaster (RCND 2019) IOP Publishing
IOP Conf. Series: Earth and Environmental Science 479 (2020) 012039 doi:10.1088/1755-1315/479/1/012039

INPUT
Radial Basis FORECASTING PROCESS OUTPUT
Function NN RBFoutput
Model

River Water
Flow
ANFIS Model ANFISoutput
Input data

Individual
Data Expert
River Water final forecasting
PreProc Forecast
Level output
essing Output
Support Vector
SVMoutput
Machine Model

Rainfall

CMIS Model Analysis

Deep Neural
DLoutput
Network Model

Figure 2. The schematic of the proposed CMIS flood disaster forecasting frameworks

The applications in flood forecasting can be classified according to flood resource variables, i.e.,
water level, flood peak discharge, urban flood, plain flood, river flood, precipitation, river inflow, peak
flow, river flow, rainfall-runoff, flash flood, rainfall, streamflow, seasonal streamflow, soil moisture,
rainfall–discharge, groundwater level, rainfall stage, flood frequency analysis, flood quantiles, surge
level, extreme flow, storm surge, typhoon rainfall, and daily flows [23]. Among these key influencing
flood resource variables, rainfall and the streamflow river water level had the most remarkable role in
flood modelling [24]. In Figure 2, three differences input data includes river water flow (streamflow),
river water level, and rainfall is proposed in this study.

3.1. Ensemble Model Based on CMIS

Committee machines attempt to minimize the errors of individual learning algorithms or machines by
grouping them and making them work synergistically. The ensemble is a more robust model than the
model represented by any individual machine. The last phase of designing committee machine with
intelligent system is the combination of the individual-intelligent outputs. Numerous examinations have
been done to discover the consolidation techniques to combine the individual outputs and produce the
final output values. In committee machine methods, the ensemble candidates are different. There are a
number of methods to create different individual training data, the initial condition, the topology of nets,
and the training algorithms. After selecting individuals and training them, their generated results will be
combined with some methods.
There are two methods to determine weights for CMIS; simple ensemble averaging using equal
weights and weighted averaging using optimized weights [39]. In the simple ensemble averaging
approach, the outputs can just take the average as given by Equation (1), and the weighted averaging
approach with the gates ( ( can be generated in any convenient manner, the outputs are gated
according to the inputs.

1
( = ( ( = ( ( ( (1)

An ensemble average consists of a set of training models which share a common input ( for
training pattern , and whose individual outputs ( are combined to produce an overall output ( .
The authors Opitz and Shavlik [42] presented the algorithm that uses genetic algorithm to explicitly
search for a highly diverse set of accurately trained networks. Application on the permeability prediction

7
The 7th AUN/SEED-Net Regional Conference on Natural Disaster (RCND 2019) IOP Publishing
IOP Conf. Series: Earth and Environmental Science 479 (2020) 012039 doi:10.1088/1755-1315/479/1/012039

using committee machine also presented in [17], ensemble averaging method is computed according to
the weight. The optimal combination of the weight for prediction is also investigated using genetic
algorithm [43]. The proposed combining method using fuzzy genetic algorithm gives smallest error and
highest correlation on the reliability of the permeability predictions [41]. While the authors [44] obtained
optimal weight factors by using a genetic algorithm-pattern search (GA-PS) to predict Poisson’s ratio.
The model constructed by CMIS approach consists of radial basis neural network, Sugeno fuzzy
inference system and ANFIS models.
To obtain the optimal weights for combining using GA algorithm, the fitness function as defined in
the Equation (2).

1
= ( + +⋯ − ; =1 (2)
=1 =1

Where, is the output of the first network on the − ℎ input or − ℎ training pattern, is the
weight of the ℎ member, is the target value of ℎ input, and is the number of training data.

4. Conclusion
The present formulation of developed ML modelling for flood forecasting is relatively young and in the
early stage of advancement. Previous studies indicated that in the context for improving the quality of
prediction, the novel ensemble and advance hybridizations methods are challenging to be studied. An
applicable flood forecasting model using advance committee machine learning approach is proposed in
this work. This new understanding should help to improve the forecasting technique particularly in flood
disasters. In general, the proposed CMIS framework is expected to exhibit itself as very optimistic
predictive model that can be utilized as a viable alternative to the state-of-the-art of advanced soft
computing for flood forecasting technique.
Further work should be undertaken to develop individual machine learning algorithms, constructing
the committee machine with intelligent systems model and employs these models into some considered
case study. Enhance the CMIS model and algorithm to get a better result, more robustness flood
forecasting model and reliable design. Examine the proposed models into the real system in term of
benchmarking study can be considered. The national flood prevention and warning program as known
as Program Ramalan dan Amaran Banjir Negara, Malaysia, therefore can be an option in case of
benchmarking application. The potential proposed method can also be tested in diverse area flood
forecasting data in a way to check the generalization capability of the CMIS model.

Acknowledgements
The authors would like to acknowledge the Malaysia-Japan International Institute of Technology,
Universiti Teknologi Malaysia, for supporting and facilitating this work. The first author is grateful to
Universitas Muhammadiyah Malang for supporting his study in undertaking this work. The sponsorship
given by Japan-ASEAN Integration Fund (JAIF) is very much appreciated.

References
[1] M. S. Khalid and S. Shafiai, “Flood Disaster Management in Malaysia: An Evaluation of the
Effectiveness Flood Delivery System,” Int. J. Soc. Sci. Humanit., vol. 5, no. 4, pp. 398–402,
2015.
[2] S. K. Jain et al., “A Brief review of flood forecasting techniques and their applications,” Int. J.
River Basin Manag., vol. 16, no. 3, pp. 329–344, 2018.
[3] UNISDR, “Sendai Framework for Disaster Risk Reduction 2015-2030,” 2015.
[4] A. K. Lohani, N. K. Goel, and K. K. S. Bhatia, “Improving real time flood forecasting using
fuzzy inference system,” J. Hydrol., vol. 509, pp. 25–41, 2014.
[5] T. Zhao et al., “Statistical and Hybrid Methods Implemented in a Web Application for Predicting
Reservoir Inflows during Flood Events,” J. Am. Water Resour. Assoc., vol. 54, no. 1, pp. 69–89,

8
The 7th AUN/SEED-Net Regional Conference on Natural Disaster (RCND 2019) IOP Publishing
IOP Conf. Series: Earth and Environmental Science 479 (2020) 012039 doi:10.1088/1755-1315/479/1/012039

2018.
[6] G. Napolitano, L. See, B. Calvo, F. Savi, and A. Heppenstall, “A conceptual and neural network
model for real-time flood forecasting of the Tiber River in Rome,” Phys. Chem. Earth, vol. 35,
pp. 187–194, 2010.
[7] S. Zhu, J. Zhou, L. Ye, and C. Meng, “Streamflow estimation by support vector machine coupled
with different methods of time series decomposition in the upper reaches of Yangtze River,
China,” Environ. Earth Sci., vol. 75, no. 6, pp. 1–12, 2016.
[8] M. Ashrafi, L. H. C. Chua, C. Quek, and X. Qin, “A fully-online Neuro-Fuzzy model for flow
forecasting in basins with limited data,” J. Hydrol., vol. 545, pp. 424–435, 2017.
[9] Z. Guo, K. Zhou, X. Zhang, and S. Yang, “A deep learning model for short-term power load and
probability density forecasting,” Energy, vol. 160, pp. 1186–1200, 2018.
[10] L. Qu, W. Li, W. Li, D. Ma, and Y. Wang, “Daily long-term traffic flow forecasting based on a
deep neural network,” Expert Syst. Appl., vol. 121, pp. 304–312, 2019.
[11] X. He, J. Luo, G. Zuo, and J. Xie, “Daily Runoff Forecasting Using a Hybrid Model Based on
Variational Mode Decomposition and Deep Neural Networks,” Water Resour. Manag., 2019.
[12] M. Cai, M. Pipattanasomporn, and S. Rahman, “Day-ahead building-level load forecasts using
deep learning vs. traditional time-series techniques,” Appl. Energy, vol. 236, pp. 1078–1088,
2019.
[13] R. J. Abrahart and L. See, “Multi-model data fusion for river flow forecasting: an evaluation of
six alternative methods based on two contrasting catchments,” Hydrol. Earth Syst. Sci., vol. 6,
no. 4, pp. 655–670, 2002.
[14] M. Goswami and K. M. O’Connor, “Real-time flow forecasting in the absence of quantitative
precipitation forecasts: A multi-model approach,” J. Hydrol., vol. 334, no. 1–2, pp. 125–140,
2007.
[15] P. Bagheripour, “Committee neural network model for rock permeability prediction,” J. Appl.
Geophys., vol. 104, pp. 142–148, 2014.
[16] R. Prasad, R. C. Deo, Y. Li, and T. Maraseni, “Ensemble committee-based data intelligent
approach for generating soil moisture forecasts with multivariate hydro-meteorological
predictors,” Soil Tillage Res., vol. 181, pp. 63–81, 2018.
[17] C. H. Chen and Z. S. Lin, “A committee machine with empirical formulas for permeability
prediction,” Comput. Geosci., vol. 32, no. 4, pp. 485–496, 2006.
[18] A. Kadkhodaie-Ilkhchi, H. Rahimpour-Bonab, and M. Rezaee, “A committee machine with
intelligent systems for estimation of total organic carbon content from petrophysical data: An
example from Kangan and Dalan reservoirs in South Pars Gas Field, Iran,” Comput. Geosci., vol.
35, no. 3, pp. 459–474, 2009.
[19] A. Gholami, H. R. Ansari, and S. Ahmadi, “Combining of intelligent models through committee
machine for estimation of wax deposition,” J. Chinese Chem. Soc., vol. 65, no. 8, pp. 925–931,
2018.
[20] A. Mosavi, P. Ozturk, and K. W. Chau, “Flood prediction using machine learning models:
Literature review,” Water (Switzerland), vol. 10, no. 11, pp. 1–40, 2018.
[21] F. Fotovatikhah, M. Herrera, S. Shamshirband, K. W. Chau, S. F. Ardabili, and M. J. Piran,
“Survey of computational intelligence as basis to big flood management: Challenges, research
directions and future work,” Eng. Appl. Comput. Fluid Mech., vol. 12, no. 1, pp. 411–437, 2018.
[22] Z. M. Yaseen, S. O. Sulaiman, R. C. Deo, and K. W. Chau, “An enhanced extreme learning
machine model for river flow forecasting: State-of-the-art, practical applications in water
resource engineering area and future research direction,” J. Hydrol., vol. 569, no. November
2018, pp. 387–408, 2019.
[23] H. R. Maier, A. Jain, G. C. Dandy, and K. P. Sudheer, “Methods used for the development of
neural networks for the prediction of water resource variables in river systems: Current status
and future directions,” Environ. Model. Softw., vol. 25, no. 8, pp. 891–909, 2010.
[24] M. Toukourou, A. Johannet, G. Dreyfus, and P. A. Ayral, “Rainfall-runoff modeling of flash
floods in the absence of rainfall forecasts: The case of ‘cévenol flash floods,’” Appl. Intell., vol.
35, no. 2, pp. 178–189, 2011.

9
The 7th AUN/SEED-Net Regional Conference on Natural Disaster (RCND 2019) IOP Publishing
IOP Conf. Series: Earth and Environmental Science 479 (2020) 012039 doi:10.1088/1755-1315/479/1/012039

[25] E. Álvarez-Fanjul, B. Pérez, and I. Rodríguez, “Nivmar: a storm surge forecasting system for
Spanish waters,” Sci. Mar., vol. 65, pp. 145–154, 2001.
[26] A. Jabbari and D. H. Bae, “Application of Artificial Neural Networks for accuracy enhancements
of real-time flood forecasting in the Imjin basin,” Water (Switzerland), vol. 10, no. 11, 2018.
[27] P. K. T. Nguyen, L. H. C. Chua, and L. H. Son, “Flood forecasting in large rivers with data-
driven models,” Nat. Hazards, vol. 71, no. 1, pp. 767–784, 2014.
[28] L. H. Feng and J. Lu, “The practical research on flood forecasting based on artificial neural
networks,” Expert Syst. Appl., vol. 37, no. 4, pp. 2974–2977, 2010.
[29] G. Tayfur, V. P. Singh, T. Moramarco, and S. Barbetta, “Flood hydrograph prediction using
machine learning methods,” Water (Switzerland), vol. 10, no. 8, pp. 1–13, 2018.
[30] C. Hu, Q. Wu, H. Li, S. Jian, N. Li, and Z. Lou, “Deep learning with a long short-term memory
networks approach for rainfall-runoff simulation,” Water, vol. 10, no. 11, pp. 1–16, 2018.
[31] B. Wang, L. Zheng, and D. L. Liu, “Using multi-model ensembles of CMIP5 global climate
models to reproduce observed monthly rainfall and temperature with machine learning methods
in Australia,” Int. J. Clim., no. May, pp. 1–12, 2018.
[32] H. Shafizadeh-Moghadam, R. Valavi, H. Shahabi, K. Chapi, and A. Shirzadi, “Novel forecasting
approaches using combination of machine learning and statistical models for flood susceptibility
mapping,” J. Environ. Manage., vol. 217, pp. 1–11, 2018.
[33] A. Tatar, M. R. Yassin, M. Rezaee, A. H. Aghajafari, and A. Shokrollahi, “Applying a robust
solution based on expert systems and GA evolutionary algorithm for prognosticating residual gas
saturation in water drive gas reservoirs,” J. Nat. Gas Sci. Eng., vol. 21, pp. 79–94, 2014.
[34] M. Systems, “Prediction of Stock Market Using an Ensemble Learning-based Intelligent Model,”
Ind. Eng. Manag. Syst., vol. 17, no. 3, pp. 479–496, 2018.
[35] A. Y. Shamseldin and K. M. O. Connor, “A " consensus " real-time river flow forecasting model
for the Blue Nile River,” in Proceedings of Symposium Water Resources Systems-Hydrological
Risk, Management and Development, 2003, no. 28, pp. 82–89.
[36] A. Dashti, H. Riasat, and M. Rezakazemi, “Accurate prediction of solubility of gases within H 2
-selective nanocomposite membranes using committee machine intelligent system,” Int. J.
Hydrogen Energy, vol. 43, pp. 6614–6624, 2018.
[37] N. Ghadimi, A. Akbarimajd, H. Shayeghi, and O. Abedinia, “A new prediction model based on
multi-block forecast engine in smart grid,” J. Ambient Intell. Humaniz. Comput., vol. 9, pp.
1873–1888, 2018.
[38] M. Azmi, S. Araginejad, and M. Kholighi, “Multi Model Data Fusion for Hydrological
Forecasting using K-Nearest Neighbour Method,” Iran. J. Sci. Technol., vol. 34, no. B1, pp. 81–
92, 2010.
[39] A. A. Nadiri, E. Fijani, F. T.-C. Tsai, and A. Asghari Moghaddam, “Supervised committee
machine with artificial intelligence for prediction of fluoride concentration,” J.
Hydroinformatics, vol. 15, no. 4, pp. 1474–1490, 2015.
[40] S. Haykin, Neural Networks: A Comprehensive Foundation, Second Edi. Pearson Prentice Hall,
1999.
[41] S. A. Jafari, S. Mashohor, and M. J. Varnamkhasti, “Committee neural networks with fuzzy
genetic algorithm,” J. Pet. Sci. Eng., vol. 76, pp. 217–223, 2011.
[42] D. W. Opitz and J. W. Shavlik, “Actively searching for an effective neural-network ensemble,”
Conn. Sci., vol. 8, no. 3, p. 337—354, 1996.
[43] A. Rostami, M. Kalantari-Meybodi, M. Karimi, A. Tatar, and A. H. Mohammadi, “Efficient
estimation of hydrolyzed polyacrylamide (HPAM) solution viscosity for enhanced oil recovery
process by polymer flooding,” Oil Gas Sci. Technol., vol. 73, no. 22, pp. 1–17, 2018.
[44] M. Asoodeh, “Prediction of poisson’s ratio from conventional well log data: A committee
machine with intelligent systems approach,” Energy Sources, Part A Recover. Util. Environ. Eff.,
vol. 35, no. 10, pp. 962–975, 2013.

10