Natural Disaster Application on Big Data and

Machine Learning: A Review

1st Rania Rizki Arinta 2nd Andi W.R. Emanuel

Magister Teknik Informatika Magister Teknik Informatika
Universitas Atma Jaya Yogyakarta Universitas Atma Jaya Yogyakarta
Yogyakarta, Indonesia Yogyakarta, Indonesia
[email protected] [email protected]

Abstract— Natural disasters are events that are difficult to pattern obtained from data analysis and the creation of hidden
avoid. There are several ways of reducing the risks of natural information [5].
disasters. One of them is implementing disaster reduction
programs. There are already several developed countries that In this study, we will review and focus more on Disaster
apply the concept of disaster reduction. In addition to disaster management phases, which focus more on the use of machine
reduction programs, there are several ways to predict or learning in the area of Disaster management phases, which
reducing the risks using artificial intelligence technology. One of consists of the data source used and the model/algorithm used.
them is big data, machine learning, and deep learning. By Because to find out whether the previous research solved the
utilizing this method at the moment, it facilitates tasks in problem in the prediction area and early detection we must
visualizing, analyzing, and predicting natural disaster. This know the data source used already has 5'v characteristics,
research will focus on conducting a review process and namely Velocity, Volume, Value, Variety, and Veracity. The
understanding the purpose of machine learning and big data in performance level of the model made is good or not from the
the area of disaster management and natural disaster. The level of accuracy, precision, recall, and the execution time.
result of this paper is providing insight and the use of big data, The propose of this study to give an insight and the use of big
machine learning, and deep learning in 6 disaster management data, machine learning, and deep learning from 6 disaster area
area. This 6-disaster management area includes early warning which is early warning damage, damage assessment,
damage, damage assessment, monitoring and detection,
monitoring and detection, forecasting and predicting, and
forecasting and predicting, and post-disaster coordination, and
post-disaster coordination, and response, and long-term risk
response, and long-term risk assessment and reduction.
assessment and reduction. But also give the data source that
Keywords— natural disaster, review, big data, machine the previous research
learning

I. INTRODUCTION II. RELATED WORK

Natural disasters are events that result from natural Based on existing research, Manzhu Yu researches on
processes that cannot be predicted. Natural disasters can also natural disasters which are divided into several parts. One of
cause loss of life or damage property and economic losses. them is categorizing the articles based on major data sources
According to data provided by world health organizations which consist of satellite imagery, aerial imagery, and videos
from 1900 to 2018, there are around 14 million for all types of from unmanned Detection and Ranging (LiDAR), simulation,
disasters [1]. Children are the most vulnerable groups who get spatial data, crowdsourcing, social media, and mobile.
direct impacts of disasters. Disaster victims come from Disaster management phases consist of 4 main parts
various countries, and disasters have been threatening the lives Mitigation/prevention, Preparedness, Response, and
of millions of children [2]. Recovery. But in Manzhu’s research, they divided the 4 parts
One effort to minimize disasters is to provide a program of the Disaster Management Phases into 6 phases, which
for reducing the risks of natural disasters. Disaster risk include early warning damage, damage assessment,
reduction is a concept of how communities reduce damage monitoring and detection, forecasting and predicting, and
and victims affected by disasters. One example of risk post-disaster coordination, and response, and long-term risk
reduction according to the Sendai framework [3] made by assessment and reduction. Based on Goswami's research, he
UNISDR (United Nations International Strategy for Disaster conducted a review of natural disasters in a case study in India.
Reduction) is understanding the risk of the disaster. According This research classifies the objectives of the tasks, namely
to Goswami's research [4] states that the aim of disaster Prediction, Detection, and Disaster management strategies.
management is minimizing victims, can save victims Based on this data analysis, we could determine the pattern
promptly, evacuate people to save places, Reconstruct the and solve the problem. By predicting the event or maybe use
damages immediately, Offer first aid instantly. it for early detection. To solve this kind of problem, we need
The technology today is big data, machine learning, along to use not only data sources but also the type of model/
with deep learning. According to Hashem's research stated algorithm to train and test the data. Most of the review
that big data is a set of techniques and technology that requires focuses on the only area on the disaster management phase
a new form of integration to find large hidden values derived and the data source from the result of Manzhu study he didn’t
from datasets that are complex, diverse, and of a massive explain about the performance on the model/algorithm in the
scale. The advantage of using big data can determine the reviews. The most important part of the early detection and

XXX-X-XXXX-XXXX-X/XX/$XX.00 ©20XX IEEE

Authorized licensed use limited to: Bahria University. Downloaded on November 11,2024 at 08:55:07 UTC from IEEE Xplore. Restrictions apply.
 !"" !"#!"$$%&'(  249
                        
   

prediction is to analyze the data source and the model they

use to solve the problem.
The following table 1 is a table of previous researchers that
focuses on the area topic in disaster management and natural
disaster using big data and machine learning approach.

TABLE I. RELATED WORK ON THE PREVIOUS REVIEW RESEARCH

PAPER

Categories Data source Citation

• Early warning • Satellite imagery [6]
damage • Aerial imagery and
• Damage assessment videos from
• Monitoring and unmanned Detection
detection and Ranging
• Forecasting and (LiDAR) Fig. 1. Step Of Experimental Method
predicting • Simulation, spatial
• Post-disaster data In figure 1 it explains the steps to getting the journal so
coordination and • Crowdsourcing
that it is first selected based on the appropriate topic. Paper
response • Social media
• Long-term risk
taken must be in the area of big data, machine learning, and
assessment and deep learning. The chosen paper must have data source
reduction information used in the study. Along with the
model/technique/algorithm used. In this paper, we will
• Prediction, • Hydrological data, [4] classify reviews based on frequently used data sources, often
• Detection • Meteorological data used algorithms so that it can compare and provide evaluations
• Disaster • twitter of datasets that are accurately used in big data research,
management
strategies machine learning, or deep learning. Besides that, it categorizes
Based on previous research such as the table above in the based on Disaster management phases that have been
study, Manzhu Yu divides into 6 categories, and Goswami modified by human management, which consists of 6
categories [6]. There are 6, namely early warning damage,
divides into 3 groups. Based on these categories, Manzhu's
damage assessment, monitoring and detection, forecasting and
research was more detailed because the category was taken predicting post-disaster coordination, and response, and long-
based on the concept of Disaster management phases. In this term risk assessment and reduction. Its long-term risk
study, we will adopt 6 categories that have been created by assessment and reduction function is for long-term disaster
Manzhu, and combine them with Goswami research to risk reduction. Forecasting and Predicting used to predict
provide models/techniques that have been given along with natural disasters that will occur. Monitoring and detection to
the data used. detect or monitor disasters. Early warning Damage is used to
receive relevant and timely information. Damage Assessment
III. METHOD is an actual disaster evaluation event caused by a natural
The paper that will be chosen in the first review process is disaster event. Post-disaster recovery planning is defined as
a paper that publishes on international proceedings or an developing a set of strategies to help rebuild after a disaster
article published in international journals. So from that, the occurs
article to be taken must use English. After that, the second
requirement is the paper chosen is a paper published in the last IV. RESULT
five years. That is a paper published from 2014 until 2019. After choosing the paper that matches the criteria above,
After that, the article was taken based on science direct[7], then it is to group the article based on the categories that have
springer open-source [8], IEEE [9], google scholar [10], and been made by Manzhu. After that, classify the review paper
research gate [11]. The search process for the paper by on the data source used during natural disaster research. After
considering the abstract keyword is related to any natural that, it is classifying review paper based on the model/
disaster. The article must also discuss the fields of big data, algorithm used in the study as with the Goswami study. In
machine learning, or deep learning. The following figure 1 Manzhu’s study, there are six categories of Disaster
explains the flow chart on how to choose the article that will management phases. These categories are, among others,
be reviewed in this study. Long-term risk assessment and reduction, Forecasting and
Predicting, Monitoring and detection, Early Warning
Damage, Damage Assessment, Post-disaster Coordination,
and Response. The following are the results of the paper
following the six categories in Table 2.

TABLE II. DISASTER MANAGEMENT PHASES

Categories Citation
Long-term risk assessment and [12],[13],[14],[15][16]
reduction
Forecasting and Predicting [17],[18],[19],[20],[21][22][23][2
4][25][26]
Monitoring and detection [27],[28],[29],[30],[31],[32],[33],[
34],[35],[36][37][38][39][40][41]
[42][43][44][45][46]

Authorized licensed use limited to: Bahria University. Downloaded on November 11,2024 at 08:55:07 UTC from IEEE Xplore. Restrictions apply.
250
                        
   

Early warning Damage [47],[48],[49],[50][51][52][53][54 to obtain data sources such as Wieland's research [32]. Base
] on the results in table 3 concluded that the data source that is
Damage Assessment [55],[15],[56],[15],[55],[56][57][5
8] often used is social media data, especially Twitter. social
Post-disaster Coordination and [59],[60][61][62][63] media data, for example, twitter to access data on tweets is
Response accessible. Based on Nguyen’s research, he uses steaming
API for extracting the tweets data with a keyword “quake,”
Based on the results of table 2, part of the research is often “tsunami,” and “earthquake.” The study also explains that
done in the area of monitoring and detection. The monitoring they could download 18gb twitter bases on “quake,”
and detection section is a component that is usually “tsunami,” and “earthquake” keyword from September 11 to
researched because based on the results of the paper to obtain December 25, 2016 [36]. Most of the satellite data,
data sources in the process of monitoring and detection is crowdsourcing, and areal image took at least two years to
straightforward. Data sources that are often used are twitter generate the data to understand and predict the pattern from
data and satellite data. The data retrieval process is easy to the image. For example, in Thibaut’s research, he uses data
take because it is open source data so that the research is a crowdsourcing from the Oklahoma Geological Survey
research that is often carried out in the area of a natural (OGS), which only generate data 2021 seismic events from
disaster. 2014 to 2016 [29]. We could conclude that social media data
Research that is rarely done is post-disaster coordination and which is twitter data is easy to extract with the help of
response. The big data and machine learning in this area are streaming API, not only it’s easy but the volume of data is
still very uncommon. Because to conduct this area of topics, enormous we could get 18gb data on a tweet with a small
the expert has to actually examine the disaster area and carry period of time which in this case it’s 4 month rather using
out the infrastructure process from the damaged areas from crowdsourcing data which take approximately 2 years. After
natural disasters. Based on the results of table 2, it can be classifying the paper based on the data source, then classify
concluded that the most suitable area for disaster risk in the the article based on the model / technique / algorithm used in
natural disaster area is Monitoring and detection and the study. There are several types of models used. Table 4
Forecasting and Predicting. the area is the most appropriate discusses the model/algorithm used for the disaster area.
area to be applied in a natural disaster area because of the
domain is the most suitable area to be used in the area of TABLE IV. MODEL / TECHNIQUES USED

natural disasters because the data provided has already had Model/techniques Citation
5'v characteristics, namely Velocity, Volume, Value, Variety, SVSA (Support Vector Selection [28]
and Veracity. Adaptation)
Decision Tree [34][63]
Random Forest [35][65][69][34][13][61]
After categorizing according to the 6 categories, then classify Support Vector Machine [65][67][32][49][28]
based on data sources used in research on big data, machine [34],[13][16][39][58][51][61][42]
learning, and deep learning. The following data source is used [25][62][26]
for research in table 3. Boosted Regression Trees (BRT) [35]
Bayesian Networks [14]
TABLE III. TABLE DATA SOURCE FOR THE DISASTER AREA K-Means [12][39]
Naïve Bayes [51][61][43]
Data source Citation Naïve Bayes Classifiers [66][69][49][57]
Satellite imagery [35][30][64][32][28][42][26][45] CNN (Convolutional Neural [55][36][29][38][44][45][54]
Social media [47][65][66][56][55][15][67][21][ Network)
49] Latent Dirichlet Allocation (LDA) [15][37][61][41]
[31][36][37][16][57][23][61][52][ ANN (artificial neural networks) [27][17][20][70][22][42]
41][62][53][46][54][63] C4.5 [51][63]
Crowdsourcing data [65][17][21][20][29][16][22][39][ Neural Clouds [40]
58][24][25][44] Logistic Regression [42]
Aerial imagery [59][38]

Online News report [51][43] Based on the results of table 4, there are several types of
algorithms used. For the algorithm, Support Vector Machine
In table 3, there is a data source used by previous research. is used in research to detect changes in single- and multi-
The data sources are satellite imagery, social media, temporal X- and L-bands using SAR (Synthetic Aperture
crowdsourcing data, and aerial imagery. In the above results, Radar) image [32]. Besides that, it is also used in the research
the data source that is often used is social media, especially of earthquake image classification process. In this study,
Twitter. For data sources that use Twitter more, they use using SVM to compare whether the SVSA model (Support
analysis sentiment, as Ghazaleh's research uses data source Vector Selection Adaptation) can provide increased
social media for sentiment analysis [68]. Zahra's research performance at the computational time, along with the new
uses twitter data sources such as Ghazaleh's research for model does not require kernels [28]. Also, the model that is
analysis sentiment [60]. On crowdsourcing data that is often often used is ANN (artificial neural network). The study used
taken is data provided by USGS (United States Geological ANN for earthquake early detection, also known as precursor
Survey). One paper that uses these data is Mendoza’s [27]. Also, ANN was also used to predict earthquakes
research [21] and Resch's research [15].The satellite image is magnitude in Tokyo [20]. Base on the result, we conclude that
also used as a data source. The satellite image came from CNN works best for earthquake detection. Not only for the
NASA like the Gokaraju study [30]. Besides that, there is accuracy, but recall, precision, and the execution time is
satellite imagery that uses SAR (Synthetic Aperture Radar) much faster. But the downside of this algorithm/model you

Authorized licensed use limited to: Bahria University. Downloaded on November 11,2024 at 08:55:07 UTC from IEEE Xplore. Restrictions apply.
251
                        
   

need a lot of data. For example, Perol’s research uses CNN has 5'v characteristics, namely Velocity, Volume, Value,
for earthquake detection. He compares it with the previous Variety, and Veracity.
study and that his model execution time only took 1 minute 1 Base on the result of tables 2,3, and 4 we could conclude that
second then the other 2 models, which took 9 days 13 hours big data and machine learning could help overcome the
and 48 hours to execute [29]. Not only in Perol’s research but disaster management area which includes early warning
also Nguyen uses CNN for earthquake detection perform damage, damage assessment, monitoring and detection,
better not only in the accuracy, but also recall, pre, and score. forecasting and predicting post-disaster coordination, and
They compare the CNN model with SVM. The reason why response, and long-term risk assessment and reduction. From
CNN performs better than SVM it’s because of the amount of this area, we could choose the right time of data source and
data they use. Base on that research, they use 18gb tweets of model which one performs better for the set of problems that
data. SVM performs poorly due to the amount of data. SVM we choose.
works really well on smaller data rather than big amounts of
data. They use the CNN model to avoid underfitting and V. CONCLUSION
overfitting problem [36]. The neural cloud is an algorithm Based on previous results, it can be concluded that with
that uses the combination of an Advanced K-Means (AKM) big data, machine learning, or deep learning can help in 6 areas
clustering algorithm and an extended Radial Basis Functions of Disaster management. These areas are early warning
(RBF). The neural cloud is a cloud detection algorithm. damage, damage assessment, monitoring and detection,
Research by Pyayt et al. uses the algorithm for monitoring forecasting and predicting post-disaster coordination, and
and flood protection [40]. response, and long-term risk assessment and reduction. By
classifying based on data sources and models/algorithms that
Based on the result of table 2 we could see the result from are often used by previous researchers can help researchers
figure 2. who will research in the field of a natural disaster from the
result we can conclude that the most common area that utilizes
6 disaster area response big data and machine learning Monitoring and detection.
Monitoring and detection is the most common research in the
6 disaster management area due to the increase of data
collection. The data collection on monitoring and detection is
8.9% 8.9% easy to extract with twitter data. And the amount of data that
had been extracted with twitter API was 18gb twitter data.
14.3% 17.9%
And the data has 5'v characteristics, namely Velocity,
Volume, Value, Variety, and Veracity. For the future, we will
examine the model/ algorithm not only base on the level of
14.3%
accuracy, precision, recall, and the execution time but also the
time needed to build the model itself.
35.7%

REFERENCES
[1] “Number of reported disasters by type.” [Online]. Available:
Long-term risk assessment and reduction https://ourworldindata.org/natural-disasters.
Forecasting and Predicting [2] Tuswadi and T. Hayashi, “Disaster Prevention Education in
Merapi Volcano Area Primary Schools: Focusing on Students’
Monitoring and detection Perception and Teachers’ Performance,” Procedia Environ. Sci.,
vol. 20, pp. 668–677, 2014.
Early warning Damage [3] “2015_43291_Sendaiframeworkfordrren_Disaster Reducton
2015-2030,” 2015.
Damage Assessment
[4] S. Goswami, S. Chakraborty, S. Ghosh, A. Chakrabarti, and B.
Post-disaster Coordination and Response Chakraborty, “A review on application of data mining techniques
to combat natural disasters,” Ain Shams Eng. J., vol. 9, no. 3, pp.
365–378, 2018.
Fig. 2. Result Chart
[5] I. A. T. Hashem, I. Yaqoob, N. B. Anuar, S. Mokhtar, A. Gani, and
S. Ullah Khan, “The rise of ‘big data’ on cloud computing: Review
Figure 2 explains the most common research and the most and open research issues,” Inf. Syst., vol. 47, pp. 98–115, 2015.
uncommon research regarding the 6-disaster area. From the [6] M. Yu, C. Yang, and Y. Li, “Big Data in Natural Disaster
result we conclude that for the 6-disaster area the most Management: A Review,” Geosciences, vol. 8, no. 5, p. 165, 2018.
common area was monitoring and detection. The reason why [7] P. Sciences, “science direct,” 2018. [Online]. Available:
this area is so common because monitoring a disaster has a big https://www.sciencedirect.com/.
impact on the disaster management phase because if we could [8] Springer, “springeropen,” Technolo, 2018. [Online]. Available:
monitor the activities of the natural disaster, we could https://www.springeropen.com/journals.
recognize the pattern and eventually we could predict the [9] IEEE, “IEEE,” 2018. [Online]. Available:
cause of the natural disaster. From the data source area, the http://ieeexplore.ieee.org.
most common research was social media data. The reason why [10] Google Scholar, “Google Scholar,” 2018. [Online]. Available:
it was so common because the data extraction only needed a https://scholar.google.com/intl/en/scholar/about.html.
twitter API and the data that the twitter produces up to 18 GB [11] L. Connect, “ResearchGate,” 2018. [Online]. Available:
https://www.researchgate.net/.
of twitter data. The most important part of big data and
[12] B. Supriyadi, A. P. Windarto, T. Soemartono, and . M.,
machine learning was to extract a large amount of data that “Classification of Natural Disaster Prone Areas in Indonesia using

Authorized licensed use limited to: Bahria University. Downloaded on November 11,2024 at 08:55:07 UTC from IEEE Xplore. Restrictions apply.
252
                        
   

K-Means,” Int. J. Grid Distrib. Comput., vol. 11, no. 8, pp. 87–98, AIPR 2015, no. October, 2016.
2018. [31] T. Sakaki, M. Okazaki, and Y. Matsuo, “Tweet analysis for real-
[13] C. Geiß et al., “Estimation of seismic building structural types time event detection and earthquake reporting system
using multi-sensor remote sensing and machine learning development,” IEEE Trans. Knowl. Data Eng., vol. 25, no. 4, pp.
techniques,” ISPRS J. Photogramm. Remote Sens., vol. 104, pp. 919–931, 2013.
175–188, 2015. [32] M. Wieland, W. Liu, and F. Yamazaki, “Learning change from
[14] J. Wang, X. Gu, and T. Huang, “Using Bayesian networks in Synthetic Aperture Radar images: Performance evaluation of a
analyzing powerful earthquake disaster chains,” Nat. Hazards, vol. Support Vector Machine to detect earthquake and tsunami-induced
68, no. 2, pp. 509–527, 2013. changes,” Remote Sens., vol. 8, no. 10, 2016.
[15] B. Resch, F. Usländer, and C. Havas, “Combining machine- [33] A. J. Cooner, Y. Shao, and J. B. Campbell, “Detection of urban
learning topic models and spatiotemporal analysis of social media damage using remote sensing and machine learning algorithms:
data for disaster footprint and damage assessment,” Cartogr. Revisiting the 2010 Haiti earthquake,” Remote Sens., vol. 8, no.
Geogr. Inf. Sci., vol. 45, no. 4, pp. 362–376, 2018. 10, 2016.
[16] R. Kiatpanont, U. Tanlamai, and P. Chongstitvatana, “Extraction [34] W. Li, N. Narvekar, N. Nakshatra, N. Raut, B. Sirkeci, and J. Gao,
of actionable information from crowdsourced disaster data,” J. “Seismic data classification using machine learning,” Proc. - IEEE
Emerg. Manag., vol. 14, no. 6, pp. 377–390, 2016. 4th Int. Conf. Big Data Comput. Serv. Appl. BigDataService 2018,
[17] K. M. Asim, F. Martínez-Álvarez, A. Basit, and T. Iqbal, pp. 56–63, 2018.
“Earthquake magnitude prediction in Hindukush region using [35] S. Park, J. Im, E. Jang, and J. Rhee, “Drought assessment and
machine learning techniques,” Nat. Hazards, vol. 85, no. 1, pp. monitoring through blending of multi-sensor indices using
471–486, 2017. machine learning approaches for different climate regions,” Agric.
[18] N. D. Hoang and D. T. Bui, “Predicting earthquake-induced soil For. Meteorol., vol. 216, pp. 157–169, 2016.
liquefaction based on a hybridization of kernel Fisher discriminant [36] V. Q. Nguyen, H. J. Yang, K. Kim, and A. R. Oh, “Real-Time
analysis and a least squares support vector machine: a multi- Earthquake Detection Using Convolutional Neural Network and
dataset study,” Bull. Eng. Geol. Environ., vol. 77, no. 1, pp. 191– Social Data,” Proc. - 2017 IEEE 3rd Int. Conf. Multimed. Big
204, 2018. Data, BigMM 2017, pp. 154–157, 2017.
[19] K. M. Asim, A. Idris, T. Iqbal, and F. Martínez-Álvarez, [37] M. Imran, S. Elbassuoni, C. Castillo, F. Diaz, and P. Meier,
“Earthquake prediction model using support vector regressor and “Practical extraction of disaster-relevant information from social
hybrid neural networks,” PLoS One, vol. 13, no. 7, pp. 1–22, 2018. media,” WWW 2013 Companion - Proc. 22nd Int. Conf. World
[20] G. Asencio-Cortés, F. Martínez-Álvarez, A. Troncoso, and A. Wide Web, pp. 1021–1024, 2013.
Morales-Esteban, “Medium–large earthquake magnitude [38] S. N. K. B. Amit and Y. Aoki, “Disaster detection from aerial
prediction in Tokyo with artificial neural networks,” Neural imagery with convolutional neural network,” Proc. - Int. Electron.
Comput. Appl., vol. 28, no. 5, pp. 1043–1055, 2017. Symp. Knowl. Creat. Intell. Comput. IES-KCIC 2017, vol. 2017-
[21] M. Mendoza, B. Poblete, and I. Valderrama, “Nowcasting Janua, no. July 2018, pp. 239–245, 2017.
earthquake damages with Twitter,” EPJ Data Sci., vol. 8, no. 1, [39] M. Bria, A. Baban, and D. Dumitrascu, “A Region Thesaurus
2019. Approach for High-Level Concept Detection in the Natural
[22] F. Martínez-Álvarez, A. Troncoso, H. Quintián, and E. Corchado, Disaster Domain,” An Interdiscip. J., vol. 16, no. 3, pp. 423–452,
“Rainfall prediction: A Deep Learning approach,” Lect. Notes 2013.
Comput. Sci. (including Subser. Lect. Notes Artif. Intell. Lect. [40] A. L. Pyayt, I. I. Mokhov, B. Lang, V. V. Krzhizhanovskaya, and
Notes Bioinformatics), vol. 9648, no. June, 2016. R. J. Meijer, “Machine learning methods for environmental
[23] O. Aulov and M. Halem, “Human sensor networks for improved monitoring and flood protection,” World Acad. Sci. Eng. Technol.,
modeling of natural disasters,” Proc. IEEE, vol. 100, no. 10, pp. vol. 78, no. 6, pp. 118–123, 2011.
2812–2823, 2012. [41] K. Kireyev, L. Palen, and K. Anderson, “Applications of topics
[24] A. Baldominos, E. Albacete, Y. Saez, and P. Isasi, “A scalable models to analysis of disaster-related twitter data,” NIPS Work.
machine learning online service for big data real-time analysis,” Appl. Top. Model. Text Beyond, p. , 2009.
IEEE SSCI 2014 - 2014 IEEE Symp. Ser. Comput. Intell. - CIBD [42] I. Yilmaz, “Comparison of landslide susceptibility mapping
2014 2014 IEEE Symp. Comput. Intell. Big Data, Proc., pp. 1–8, methodologies for Koyulhisar, Turkey: Conditional probability,
2015. logistic regression, artificial neural networks, and support vector
[25] X. Yu, “Disaster prediction model based on support vector machine,” Environ. Earth Sci., vol. 61, no. 4, pp. 821–836, 2010.
machine for regression and improved differential evolution,” Nat. [43] M. Scharkow, “Thematic content analysis using supervised
Hazards, vol. 85, no. 2, pp. 959–976, 2017. machine learning: An empirical evaluation using German online
[26] M. N. Jebur, B. Pradhan, and M. S. Tehrany, “Manifestation of news,” Qual. Quant., vol. 47, no. 2, pp. 761–773, 2013.
LiDAR-derived parameters in the spatial prediction of landslides [44] S. Ahmad, K. Ahmad, N. Ahmad, and N. Conci, “Convolutional
using novel ensemble evidential belief functions and support neural networks for disaster images retrieval,” CEUR Workshop
vector machine models in GIS,” IEEE J. Sel. Top. Appl. Earth Obs. Proc., vol. 1984, pp. 2016–2018, 2017.
Remote Sens., vol. 8, no. 2, pp. 674–690, 2015. [45] K. Ahmad, P. Konstantin, M. Riegler, N. Conci, and P. Holversen,
[27] B. A. S. Aji, T. H. Liong, and B. Muslim, “Detection precursor of “CNN and GAN based satellite and social media data fusion for
sumatra earthquake based on ionospheric total electron content disaster detection,” CEUR Workshop Proc., vol. 1984, no.
anomalies using N-Model Articial Neural Network,” 2017 Int. September, pp. 15–18, 2017.
Conf. Adv. Comput. Sci. Inf. Syst. ICACSIS 2017, vol. 2018-Janua, [46] M. Maldonado, D. Alulema, D. Morocho, and M. Proano, “System
pp. 269–276, 2018. for monitoring natural disasters using natural language processing
[28] G. Tas, O. K. Ersoy, and M. E. Kamas, “SUPPORT VECTOR in the social network Twitter,” Proc. - Int. Carnahan Conf. Secur.
SELECTION AND ADAPTATION FOR CLASSIFICATION OF Technol., no. October, 2017.
Istanbul Technical Univ , Informatics Institute , Istanbul , Turkey [47] D. Buscaldi and I. Hernandez-Farias, “Sentiment Analysis on
, e-mail: [email protected] Purdue University , School of Microblogs for Natural Disasters Management,” Proc. 24th Int.
Electrical and Computer Eng ., Indiana , USA , e-mail: ersoy@p,” Conf. World Wide Web - WWW ’15 Companion, pp. 1185–1188,
Computer (Long. Beach. Calif)., pp. 851–854, 2009. 2015.
[29] T. Perol, M. Gharbi, and M. Denolle, “Convolutional neural [48] G. Shalunts, G. Backfried, and K. Prinz, “Sentiment Analysis of
network for earthquake detection and location,” Sci. Adv., vol. 4, German Social Media Data for Natural Disasters,” 11th Int. Conf.
no. 2, pp. 2–10, 2018. Inf. Syst. Cris. Response Manag., no. May, pp. 752–756, 2014.
[30] B. Gokaraju, A. C. Turlapaty, D. A. Doss, R. L. King, and N. H. [49] L. Maceda, J. Llovido, and A. Satuito, “Categorization of
Younan, “Change detection analysis of tornado disaster using Earthquake-Related Tweets Using Machine Learning
conditional copulas and Data Fusion for cost-effective disaster Approaches,” no. 1, pp. 229–232, 2019.
management,” 2015 IEEE Appl. Imag. Pattern Recognit. Work. [50] B. Truong, C. Caragea, A. Squicciarini, and A. H. Tapia,

Authorized licensed use limited to: Bahria University. Downloaded on November 11,2024 at 08:55:07 UTC from IEEE Xplore. Restrictions apply.
253
                        
   

“Identifying valuable information from Twitter during natural annotated Twitter corpora for NLP of crisis-related messages,”
disasters,” Proc. ASIST Annu. Meet., vol. 51, no. 1, 2014. Proc. 10th Int. Conf. Lang. Resour. Eval. Lr. 2016, pp. 1638–1643,
[51] T. Valero, M. Gómez, and V. Pineda, “Using Machine Learning 2016.
for Extracting Information from Natural Disaster News Reports,” [62] J. R. Ragini, P. M. R. Anand, and V. Bhaskar, “Big data analytics
Comput. y Sist., vol. 13, no. 1, pp. 33–44, 2009. for disaster response and recovery through sentiment analysis,”
[52] A. H. SALANS, J. A. SCHACK, and L. N. KATZ, “Natural Int. J. Inf. Manage., vol. 42, no. May, pp. 13–24, 2018.
Language Processing to the Rescue? Extracting ‘Situational [63] M. Avvenuti, F. Del Vigna, S. Cresci, A. Marchetti, and M.
Awareness’ Tweets During Mass Emergency,” Circulation, vol. 2, Tesconi, “Pulling Information from social media in the aftermath
no. 6, pp. 900–906, 1950. of unpredictable disasters,” Proc. 2015 2nd Int. Conf. Inf.
[53] I. Varga et al., “Aid is out there: Looking for help from tweets Commun. Technol. Disaster Manag. ICT-DM 2015, no.
during a large scale disaster,” ACL 2013 - 51st Annu. Meet. Assoc. November, pp. 258–264, 2016.
Comput. Linguist. Proc. Conf., vol. 1, pp. 1619–1629, 2013. [64] Y. Liu and L. Wu, “Geological Disaster Recognition on Optical
[54] C. Caragea, A. Silvescu, and A. H. Tapia, “Identifying informative Remote Sensing Images Using Deep Learning,” Procedia Comput.
messages in disaster events using Convolutional Neural Sci., vol. 91, no. Itqm, pp. 566–575, 2016.
Networks,” Proc. Int. ISCRAM Conf., no. May, 2016. [65] A. Cobo, D. Parra, and J. Navón, “Identifying Relevant Messages
[55] D. T. Nguyen, F. Ofli, M. Imran, and P. Mitra, “Damage in a Twitter-based Citizen Channel for Natural Disaster
Assessment from Social Media Imagery Data During Disasters,” Situations,” 2015.
no. December, pp. 569–576, 2017. [66] H. Li et al., “Twitter Mining for Disaster Response: A Domain
[56] S. Cresci, M. Tesconi, A. Cimino, and F. Dell’Orletta, “A Adaptation Approach,” 12th Int. Conf. Inf. Syst. Cris. Response
Linguistically-driven Approach to Cross-Event Damage Manag. (ISCRAM 2015), p. [1-7], 2015.
Assessment of Natural Disasters from Social Media Messages,” [67] S. Cresci, A. Cimino, F. Dell’Orletta, and M. Tesconi, “Crisis
pp. 1195–1200, 2016. Mapping During Natural Disasters via Text Analysis of Social
[57] K. Zahra, F. O. Ostermann, and R. S. Purves, “Geographic Media Messages,” Lect. Notes Comput. Sci. (including Subser.
variability of Twitter usage characteristics during disaster events,” Lect. Notes Artif. Intell. Lect. Notes Bioinformatics), vol. 9419, pp.
Geo-Spatial Inf. Sci., vol. 20, no. 3, pp. 231–240, 2017. 250–258, 2015.
[58] Z. Huang, J. Zhou, L. Song, Y. Lu, and Y. Zhang, “Flood disaster [68] G. Beigi, R. Maciejewski, and H. Liu, “Sentiment Analysis and
loss comprehensive evaluation model based on optimization Ontology Engineering,” vol. 639, no. February 2016, 2016.
support vector machine,” Expert Syst. Appl., vol. 37, no. 5, pp. [69] A. Fikri Aji, I. P. E. S. Putra, P. Mursanto, and S. Yazid, “Can
3810–3814, 2010. smartphones be used to detect an earthquake? Using a machine
[59] F. Ofli et al., “Combining Human Computing and Machine learning approach to identify an earthquake event,” 8th Annu.
Learning to Make Sense of Big (Aerial) Data for Disaster IEEE Int. Syst. Conf. SysCon 2014 - Proc., pp. 72–77, 2014.
Response,” Big Data, vol. 4, no. 1, pp. 47–59, 2016. [70] G. Rajguru, Y. S. Bhadauria, and S. Mukhopadhyay, “Estimation
[60] J. Peterson and B. Argo, “Tweedr: Mining Twitter to Inform of Earthquake Source Parameters Using Machine Learning
Disaster Response,” vol. 12, no. 4, pp. 354–358, 2014. Techniques,” 2018 9th Int. Conf. Comput. Commun. Netw.
Technol. ICCCNT 2018, pp. 1–7, 2018.
[61] M. Imran, P. Mitra, and C. Castillo, “Twitter as a lifeline: Human-

Authorized licensed use limited to: Bahria University. Downloaded on November 11,2024 at 08:55:07 UTC from IEEE Xplore. Restrictions apply.
254