A machine learning-based prediction and analysis of flood affected households: A case study of floods in Bangladesh

Ganguly, Kishan Kumar; Nahar, Nadia; Hossain, B.M. Rabby

doi:10.1016/j.ijdrr.2018.12.002

Cited by 41 publications

(22 citation statements)

References 27 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…We applied RF and ANNs in this study. RFs are chosen because they have a good track record in damage modeling (e.g., Amadio et al., 2019; Ganguly et al., 2019; Schröter et al., 2018; Sieg et al., 2017; Wagenaar et al., 2017; Wagenaar et al., 2018), ANNs have also been used before in flood damage models (Amadio et al., 2019; Ganguly et al., 2019), and in this study they were selected because of their ability to extrapolate and at the same time find complex nonlinear relationships. Table I provides a comparison between the ML methods.…”

Section: Methods and Datamentioning

confidence: 99%

“…RF, an ML method developed by Breiman (2001), has been used in flood damage modeling (e.g., Amadio et al., 2019; Ganguly et al., 2019; Schröter et al., 2018; Sieg et al., 2017; Wagenaar et al., 2017; Wagenaar et al., 2018). RFs are ensembles of regression trees where the data for each tree are generated by a bootstrapping resampling method.…”

Section: Methods and Datamentioning

confidence: 99%

“…Merz, Kreibich, and Lall (2013) used a similar approach to predict flood damages at individual building level. Since then such techniques have been applied by many authors to predict all sorts of impacts from natural hazards (Amadio et al., 2019; Ganguly, Nahar, & Hossain, 2019; Carvajal et al., 2018; Mayfield et al., 2018; Nateghi, Guikema, & Quiring, 2014; Schröter et al., 2014, 2018; Sieg, Vogel, Merz, & Kreibich, 2017; Wagenaar, de Jong, & Bouwer, 2017; Wagenaar, Lüdtke, Schröter, Bouwer, & Kreibich, 2018). These data‐driven damage models often use more than one variable to predict the damage (multivariable models).…”

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

Improved Transferability of Data‐Driven Damage Models Through Sample Selection Bias Correction

Wagenaar

Hermawan

Homberg³

et al. 2020

Risk Analysis

View full text Add to dashboard Cite

Damage models for natural hazards are used for decision making on reducing and transferring risk. The damage estimates from these models depend on many variables and their complex sometimes nonlinear relationships with the damage. In recent years, data‐driven modeling techniques have been used to capture those relationships. The available data to build such models are often limited. Therefore, in practice it is usually necessary to transfer models to a different context. In this article, we show that this implies the samples used to build the model are often not fully representative for the situation where they need to be applied on, which leads to a “sample selection bias.” In this article, we enhance data‐driven damage models by applying methods, not previously applied to damage modeling, to correct for this bias before the machine learning (ML) models are trained. We demonstrate this with case studies on flooding in Europe, and typhoon wind damage in the Philippines. Two sample selection bias correction methods from the ML literature are applied and one of these methods is also adjusted to our problem. These three methods are combined with stochastic generation of synthetic damage data. We demonstrate that for both case studies, the sample selection bias correction techniques reduce model errors, especially for the mean bias error this reduction can be larger than 30%. The novel combination with stochastic data generation seems to enhance these techniques. This shows that sample selection bias correction methods are beneficial for damage model transfer.

show abstract

Section: Methods and Datamentioning

confidence: 99%

Section: Methods and Datamentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Improved Transferability of Data‐Driven Damage Models Through Sample Selection Bias Correction

Wagenaar

Hermawan

Homberg³

et al. 2020

Risk Analysis

View full text Add to dashboard Cite

show abstract

“…They also concluded that ensemble decision trees more accurately predicted damage than traditional impact models. By leveraging household-level damage assessments in Bangladesh, another flood study comparing linear regression, RF, and artificial neural networks concluded that larger households and higher education levels were associated with lower flood damage (Ganguly et al 2019). This study uses ensemble decision tree algorithms, specifically RF and SGBT, to quantitatively explore the relative role that societal factors played in the structural damage caused by Hurricane María.…”

Section: General Contextmentioning

confidence: 99%

Quantifying the Role of Vulnerability in Hurricane Damage via a Machine Learning Case Study

et al. 2021

View full text Add to dashboard Cite

Predisaster damage predictions and postdisaster damage assessments often inadequately capture the intensity and spatialtemporal complexity of natural hazard-caused damage. Accurate identification of areas with the greatest need in the wake of a disaster requires assessment of both the hazards and community vulnerabilities. This study evaluated the contribution of eight hazard and vulnerability drivers of structural damage due to Hurricane María in Puerto Rico, including wind, flood, landslide, and vulnerability measures via ensemble decision tree algorithms. Results from the algorithms indicate that vulnerability measures, including a structural vulnerability index and a social vulnerability index, were the leading predictors of damage, followed by wind, flood, and landslide measures. Therefore, it is critical to consider community vulnerabilities in damage pattern analyses and targeted, predisaster mitigation efforts.

show abstract

“…There have been several studies about the detection of river flood using machine learning algorithms. For instance, Ganguly et al [3] indicate that linear regression, in their cases, generates improved results compared to random forest and multilayer perceptron. Tehrany et al [4] use two machine learning algorithms, namely Support Vector Machine (SVM) and Decision Tree for the analysis of spatial correlations between the level of importance for detecting flood areas and flood conditioning factors such as flow accumulation, elevation, and lithology.…”

Section: Introductionmentioning

confidence: 99%

Prediction of Flood Severity Level via Processing IoT Sensor Data Using a Data Science Approach

Khan

Hussain

Alaskar

et al. 2020

IEEE Internet Things M.

View full text Add to dashboard Cite

The 'riverine flooding' is deemed a catastrophic phenomenon caused by extreme climate changes and other ecological factors (e.g., amount of sunlight), which are difficult to predict and monitor. However, the use of internet of things (IoT), various types of sensing including social sensing, 5G wireless communication and big data analysis have devised advanced tools for early prediction and management of distrust events. To this end, this paper amalgamates machine learning models and data analytics approaches along-with IoT sensor data to investigate attribute importance for the prediction of risk levels in flood. The paper presents three river levels: normal, medium and high-risk river levels for machine learning models. Performance is evaluated with varying configurations and evaluations setup including training and testing of support vector machine and random forest using principal components analysis-based dimension reduced dataset. In addition, we investigated the use of synthetic minority over-sampling technique to balance the class representations within dataset. As expected, the results indicated that a "balanced" representation of data samples achieved high accuracy (nearly 93%) when benchmarked with "imbalanced" data samples using random forest classifier 10-folds cross-validations.

show abstract

A machine learning-based prediction and analysis of flood affected households: A case study of floods in Bangladesh

Cited by 41 publications

References 27 publications

Improved Transferability of Data‐Driven Damage Models Through Sample Selection Bias Correction

Improved Transferability of Data‐Driven Damage Models Through Sample Selection Bias Correction

Quantifying the Role of Vulnerability in Hurricane Damage via a Machine Learning Case Study

Prediction of Flood Severity Level via Processing IoT Sensor Data Using a Data Science Approach

Contact Info

Product

Resources

About