Next Article in Journal
Study on the Influence of Seismic Wave Parameters on the Dynamic Response of Anti-Dip Bedding Rock Slopes under Three-Dimensional Conditions
Next Article in Special Issue
Effect of Land Use and Drainage System Changes on Urban Flood Spatial Distribution in Handan City: A Case Study
Previous Article in Journal
Transfer of Natural Radionuclides from Soil to Abu Dhabi Date Palms
Previous Article in Special Issue
Evaluation of the Predictive Performance of Regional and Global Ground Motion Predictive Equations for Shallow Active Regions in Pakistan
 
 
Search for Articles:
Title / Keyword
Author / Affiliation / Email
Journal
Article Type
 
 
Advanced Search
 
Section
Special Issue
Volume
Issue
Number
Page
 
Journals
Sustainability
Volume 14
Issue 18
10.3390/su141811329
sustainability-logo
Submit to this Journal Review for this Journal Propose a Special Issue
► Article Menu

Article Menu

share Share announcement Help format_quote Cite question_answer Discuss in SciProfiles thumb_up
...
Endorse textsms
...
Comment
first_page
settings
Order Article Reprints
Font Type:
Arial Georgia Verdana
Font Size:
Aa Aa Aa
Line Spacing:
Column Width:
Background:
Article

Predicting Factors Affecting Preparedness of Volcanic Eruption for a Sustainable Community: A Case Study in the Philippines

by
Josephine D. German
1,2,
Anak Agung Ngurah Perwira Redi
3,
Ardvin Kester S. Ong
1,*,
Yogi Tri Prasetyo
1,4 and
Vince Louis M. Sumera
5
1
School of Industrial Engineering and Engineering Management, Mapúa University, 658 Muralla St., Intramuros, Manila 1002, Philippines
2
School of Graduate Studies, Mapúa University, 658 Muralla St., Intramuros, Manila 1002, Philippines
3
Industrial Engineering Department, Sampoerna University, Jakarta 12780, Indonesia
4
Department of Industrial Engineering and Management, Yuan Ze University, 135 Yuan-Tung Road, Chung-Li 32003, Taiwan
5
Department of Civil Engineering and Geological Engineering, Mapúa University, 658 Muralla St., Intramuros, Manila 1002, Philippines
*
Author to whom correspondence should be addressed.
Sustainability 2022, 14(18), 11329; https://doi.org/10.3390/su141811329
Submission received: 31 July 2022 / Revised: 2 September 2022 / Accepted: 4 September 2022 / Published: 9 September 2022
(This article belongs to the Special Issue Geological Hazards and Risk Management)
Browse Figures
Review Reports Versions Notes

Abstract

:
Volcanic eruption activity across the world has been increasing. The recent eruption of Taal volcano and Mt. Bulusan in the Philippines affected several people due to the lack of resources, awareness, and preparedness activities. Volcanic eruption disrupts the sustainability of a community. This study assessed people’s preparedness for volcanic eruption using a machine learning ensemble. With the high accuracy of prediction from the ensemble of random forest classifier (93%) and ANN (98.86%), it was deduced that media, as a latent variable, presented as the most significant factor affecting preparedness for volcanic eruption. This was evident as the community was urged to find related information about volcanic eruption warnings from media sources. Perceived severity and vulnerability led to very high preparedness, followed by the intention to evacuate. In addition, proximity, subjective norm, and hazard knowledge for volcanic eruption significantly affected people’s preparedness. Control over individual behavior and positive attitude led to a significant effect on preparedness. It could be posited that the government’s effective mitigation and action plan would be adhered to by the people when disasters, such as volcanic eruptions, persist. With the threat of climate change, there is a need to reevaluate behavior and mitigation plans. The findings provide evidence of the community’s resilience and adoption of mitigation and preparedness for a sustainable community. The methodology provided evidence for application in assessing human behavior and prediction of factors affecting preparedness for natural disasters. Finally, the results and findings of this study could be applied and extended to other related natural disasters worldwide.

1. Introduction

Volcanic eruptions have been widely monitored and assessed for mitigation and preparation worldwide [1,2]. There are 1508 active volcanoes across 86 countries, and a population of 29 million is estimated to live within a 10-km range of volvanoes, while 800 million live within 100 km [3]. With communities living in areas of active volcanoes, the limited assessment of behavior for preparation has been widely underexplored. Niroa and Nakamura [4] explained how frameworks had been developed to assess the behavioral aspects of volcanic risks, coping mechanisms, and perception of the impact of eruption aftermath. However, with beliefs, cultural differences, and ways of governance, an assessment for community preparedness should still be considered.
Brown et al. [3] have identified several risks with volcanic eruptions, which have caused 278,368 fatalities from 1500AD to 2017. Several studies have started defining and assessing the characteristics of the population living in active volcanoes, their socio-economic status, and their perception of risk and knowledge [5]. With studies such as Reyes-Hardy et al. [6] and Barone et al. [7], it could be posited that information regarding volcanic eruption has been developed and is still developing mitigation plans. On the other hand, volcanic risk assessment for preparedness has been a trend in developing countries [2]. It was stated that community vulnerability has scarce data, and people’s behavioral intentions should be explored more [7].
In the Republic of Congo, Michellier et al. [8] assessed the community’s vulnerability to risks of a volcanic eruption. Their study focused on the population vulnerability assessment using the Social Vulnerability Index and community exposure to lava flow. Their results showed a mitigation plan using the Operational Vulnerability Index to communicate with the government for action. In Chile, Reyes-Hardy et al. [6] explored the vulnerability, volcanic hazard, and overall risk assessment using GIS-based volcanic hazard. It was seen that their model was able to assess areas of high vulnerability. However, their study focused on the physical, territorial, and social contexts. In Indonesia, Thouret et al. [5] analyzed hazard knowledge, socio-demographic characteristics, and community adaptation to volcanic threats. Their study considered hierarchical clustering (HC) to assess livelihood, demographics, and sustainability among resilient communities living in the active volcano area. However, it was evident that HC was limited to smaller datasets, user set cluster numbers, and, once done, HC cannot be returned to the original state. In Vanuatu, Niroa and Nakamura [4] assessed the indigenous disaster risk reduction framework for volcanic hazards. Their results showed that culture and belief played the most significant role in why the population would plan to mitigate the risk of a volcanic eruption. In addition, the study of Dogar and Sato [9] highlighted the consideration of the regional climate response of El Niño-Southern Oscillation (ENSO), which foregoes with volcanic eruption. They presented how ENSO phenomena occur due to volcanic explosion and that understanding of this should be considered. Due to high climate variability in the present time, all volcanic eruptions and related phenomena should be part of the discovery for human preparedness. In the study, climate variability and how it can greatly affect the phenomena of natural disasters was uncovered. Similarly, regional climate has been discussed as a sensitive aspect affecting climate trends [10]. These coincidences should be considered upon evaluation of volcanic eruptions, which lead to behavioral changes once all aspects have been considered [9,10]. Similar patterns of regional climate change were seen in the study by Dogar et al. [11]. Their study expounded on the hemispheric climate and temperature anomalies since the eruption of Mt. Pinatubo in 1991. Despite these findings, challenges in assessing behaviors for volcanic eruption preparedness were evident.
In the Philippines, Kurata et al. [2] assessed the preparation beliefs of people towards the Taal volcanic partial eruption. Their study also presented that the Philippines has been ranked third highest in the disaster risk index where active volcanoes are evidently present, as seen in Figure 1. Figure 1 represents the active and potentially active volcanoes in the Philippines [2]. With their study, several relationships were seen to be insignificant. In addition, the results showed that the causal relationship was seen to be a challenge. This is because structural equation modeling (SEM) was utilized. Fan et al. [12] expounded on the limitations of utilizing SEM as the multivariate tool to assess behaviors. It was explained that the farther the independent variable, the lower might significance be present. Moreover, the presence of mediating effects also hinders the value of the beta coefficient, representing the significance level of the independent latent variable [13]. Thus, the current studies have utilized a machine learning algorithm hybrid with SEM. Duarte and Pinho [14] have presented how the framework integration suffices within the limitations of the sole SEM methodology.
As of March 2022, the Taal volcano is still active with small earthquakes and seismic movement [15]. A short burst of 1500 m of ash means residents near the area are advised to leave the premises. Another active volcano in the Philippines, Mount Bulusan in Sorsogon, spewed volcanic ash which covered the whole town in June 2022 [16,17]. The need for evacuation among people was immediately coordinated as the community was left surprised. The World Data [18] presented that there has been a total of 44 eruptions in the past 400 years, leaving 7400 casualties. People have been seen to be unprepared for volcanic eruptions that may happen in their proximity. The integrated Theory of Planned Behavior (TPB) and Protection Motivation Theory (PMT) may be applied to assess the intention and preparedness of people for a possible volcanic eruption.
Several studies have considered utilizing extended PMT [19] or the integration of TPB with PMT [20,21,22] to assess calamities, health-related mitigation, and natural disaster preparedness. It was seen that the integrated framework could holistically assess human behavior for intention, mitigation, and preparedness for natural disasters. Gumasing et al. [19] extended the PMT to measure the effects of response efficacy of people towards typhoons. However, limitations in the measurement of the behavioral aspect was seen. Ong et al. [20] assessed the intention to prepare for mitigation of “The Big One” earthquake that is expected to happen in the Philippines, utilizing SEM. Kurata et al. [20] measured the effectiveness of response for typhoons utilizing the integrated framework with SEM. In addition, Prasetyo et al. [22] utilized the integrated framework to measure the effectiveness of community quarantine during the COVID-19 pandemic utilizing SEM. It was seen among all studies that the integrated framework would holistically measure peoples’ intentions and mitigation behavior towards health-related disasters and natural disasters. In addition, the extension by adding latent variables was effective with the framework.
The study by Kurata et al. [2] assessed people’s preparedness for the Taal volcano. This study aimed to assess the limitations and decipher the most significant contributing factor to the preparedness for a volcanic eruption in the Philippines. Several factors under PMT, such as hazard knowledge, perceived risk proximity, perceived severity, and perceived vulnerability, were considered. In addition, factors of TPB, such as perceived behavioral control, subjective norm, and attitude, were also considered to assess intention to evacuate and preparedness, with media as an extension of the integrated framework. Machine learning algorithm (MLA) ensembles, such as artificial neural network (ANN) and random forest classifier (RFC), were used to analyze the latent variables simultaneously. Similar to the study by Ong et al. [23,24], utilizing a machine learning ensemble was claimed to be more effective in analyzing factors affecting human behavior relating to the use of technology. However, no studies have utilized MLA ensemble to assess preparedness for natural disasters such as volcanic eruptions.
The present research is considered the first to analyze and predict factors affecting volcanic eruption preparedness using MLA. This study’s results would benefit researchers considering a new method of assessing human behavior worldwide. In addition, governments could utilize this study’s findings to create mitigation plans for an emergency response to volcanic eruption events worldwide. The content of the manuscript is as follows: (1) introduction, (2) conceptual framework, (3) methodology and data analysis, (4) results, (5) discussion and recommendation, and (6) conclusion.

2. Conceptual Framework

Presented in Figure 2 is the conceptual framework utilized in this study through the integration of PMT and TPB. The TPB and PMT factors are separated, as shown by broken lines in the figure. In this framework, TPB was considered as a whole since several studies [20,21,22] have shown how PMT affects TPB through Perceived Severity and Perceived Vulnerability. This study utilized an MLA ensemble, which could more effectively measure the non-linear relationship present in the framework [23,24]. A total of nine hypotheses were considered to evaluate factors affecting the preparedness for volcanic eruption of people within proximity of active volcanoes. The further building of hypotheses is discussed in this section.
Media as the primary information source was considered an extended latent variable in this study. Media was seen to affect factors under PMT, as indicated by Ong et al. [20]. Ong et al. [24] stated that the community’s primary sources of news and information would come from different media sources, such as television, newspapers and articles, and social media. The study of Kurata et al. [2] also presented how media is the main contributing factor that affects information concerning a natural disaster. Weichselgartner and Pigeon [25] also expounded on the importance of media as a source of knowledge and information regarding disasters to promote mitigation and action plans. Phengsuwan et al. [26] presented how media is a significant factor in enhancing management of disaster risks. It was seen how it delivers perception and knowledge regarding monitoring, risks, and susceptibility of natural disasters, which would lead to the intention and preparation for what may happen during, and in the aftermath of, a natural disaster. Therefore, it was hypothesized that:
H1. 
Media is a significant factor affecting Preparedness for Volcanic Eruption.
Hazard Knowledge is the information obtained, together with people’s prior experiences, regarding natural disasters. Dube and Munsaka [27] presented that knowledge from prior experience would lead to the perception of severity and vulnerability among the community when a natural disaster happens. Through patterns and recognition, it was seen that the community would be ready for natural disasters such as floods. With volcanic eruptions, it was seen that Hazard Knowledge played a significant role in the Perceived Severity and Perceived Vulnerability, which would lead to the enhanced preparedness of people [2]. If the community has a heightened perception of the risks that may affect their health, they would be more inclined to prepare for it [28]. Thus, the following were hypothesized:
H2. 
Hazard Knowledge is a significant factor affecting Perceived Severity for Preparedness for Volcanic Eruption.
H3. 
Hazard Knowledge is a significant factor affecting Perceived Vulnerability for Preparedness for Volcanic Eruption.
The proximity of the source for natural disasters plays a significant role, especially for volcanic eruptions [3]. Arias et al. [29] explained how the population living near natural disasters has a higher perception of risk (i.e., perceived severity and perceived vulnerability). Yagoub and Al Yammahi [30] justified how the spatial distribution of hazard proximity dramatically affects the severity and vulnerability of the community. Knowledge and experience affect the perceived severity and perceived vulnerability due to how close people are to the source of disasters [29]. However, it was seen that some people would perceive the risk to be lower living near the source, due to their heightened preparedness from experience. In addition, Rana et al. [31] presented how proximity to the source of disaster increases the perceived risks, such as severity and vulnerability, which, in turn, leads to an increase in preparation. Therefore, the following were hypothesized:
H4: 
Perceived Risk Proximity is a significant factor affecting Perceived Severity for Preparedness for Volcanic Eruption.
H5: 
Perceived Risk Proximity is a significant factor affecting Perceived Vulnerability for Preparedness for Volcanic Eruption.
Factors under TPB, such as Perceived Behavioral Control, Subjective Norm, and Attitude, are the main latent variables that could be used to assess behavior altogether [32]. Ong et al. [20] have presented how AT, followed by SN, and PBC consequently affected the intention to prepare for natural disasters, preceded by factors under PMT. Their study expounded on how the indirect effect of people’s understanding and knowledge led to an increase in perceived risks, affecting the intention to prepare for a natural disaster. The study by Kurata et al. [21] showed how factors of TPB were all highly significant to the perceived effectiveness of preparation for a natural disaster that may happen. In addition, Vinnell et al. [33] also considered TPB latent variables to assess preparation for natural hazards. Their study showed that distinct features of these latent variables holistically measure the community’s preparedness, highlighting how TPB alone lacks dimensions for a total measurement of behavior. As supported, this study preceded TPB with PMT to holistically measure health-related behaviors for natural disasters. Thus, it was hypothesized that:
H6. 
Perceived Behavioral Control is a significant factor affecting Preparedness for Volcanic Eruption.
H7. 
Subjective Norm is a significant factor affecting Preparedness for Volcanic Eruption.
H8. 
Attitude is a significant factor affecting Preparedness for Volcanic Eruption.
Intention to prepare has been seen to significantly and highly affect preparedness for natural disasters [20]. Najafi et al. [34] expounded on how TPB factors greatly affected intention, which led to people’s preparedness for natural disasters. The motivation of people through their intentions would significantly affect their preparedness. Bronfman et al. [35] discussed how levels of preparedness increased people’s intention, especially if the community is exposed to significant hazards from natural disasters. Bourque et al. [36] also explained how the behavioral aspect of people affected their intention and significantly increased measures of preparedness. Heller et al. [37], Kurata et al. [21], and Bourque et al. [36] explained that people’s preparedness is affected by their experience, knowledge, perception of risk, and behavioral aspect. When people know the adverse effects of hazards on health, both intention and preparation were considered directly proportional [38]. Therefore, it was hypothesized that:
H9. 
Intention is a significant factor affecting Preparedness for Volcanic Eruption.

3. Methodology

3.1. Demographics

A total of 653 valid responses were collected for this study resulting in 39,833 datasets. The data was collected from January to March 2022 through social media platforms using Google Forms with the survey questionnaire adopted from the study of Kurata et al. [2]. Through convenience sampling, no nonresponse and missing data were seen from the collected responses. Performing the Common Method Bias, 25.21% was attained, indicating no CMB in the responses [39]. The data presented 49.46% male and 50.54% female within an age range of 25–34 years old (54.52%), followed by 18–24 years old (28.02%), 35–45 years old (10.87%), and the rest were older than 45. From the responses, the majority were employed (60.16%), students (34.24%), and unemployed (5.60%) with monthly incomes within 15,000–30,000 PhP or 30,001–45,000 PhP. Most of the respondents were single and living in rural areas (50.66%) rather than urban areas (49.34%). In addition, some respondents were insured (55.67%), while 44.33% answered that they are not. A total of 79% answered that they had their own house and lot, rather than living in a condominium or renting.

3.2. Random Forest Classifier

Data pre-processing was conducted to analyze the dataset for input among the machine learning ensemble. Missing data and outliers were checked upon pre-processing and correlation analysis was conducted to determine insignificant indicators. From the correlation analysis, a threshold following the study of Ong et al. [23,24] was set with a p-value of 0.05 and coefficient of 0.20. Following which, data was aggregated using mean values representing the different factors. These data were set as the input for the machine learning ensemble run through Python 5.1.
A random forest classifier (RFC) is a classification model that considers a simple algorithm with higher prediction accuracy. RFC has been widely utilized in decision-making, natural disaster, and human behavior studies. Kim et al. [40] considered RFC to classify seismic facies. It was proven from their study that RFC produces better accuracy compared to the basic decision tree. Flood damage analysis was run using RFC by Snehil and Goel [41] and showed how an increase in prediction based on accuracy has been achieved with the simple RFC algorithm. Chen et al. [42] considered the MLA ensemble of RFC and artificial neural network (ANN) in predicting disaster risk from a flood in China. It was proven that an MLA ensemble could predict human behavior and risk assessment regarding natural disasters, specifically floods.
In addition, Yang and Zhou [43] analyzed carbon emission among residents with different tree classification techniques and presented how RFC would be primarily applicable with its capabilities to generate the best decision tree among other classifiers. Its ability to classify the optimum output from its criterion, splitter, and depth provides its advantage over the other tree classifiers. Similarly, Ong et al. [24] compared RFC and the basic DT in terms of calculation and output. It was seen that RFC dominated with a high accuracy rate. A similar study in [23] provided the pseudocode for the RFC and represented how advantageous RFC is even at the optimization stage. All possible parameters, such as the criterion, splitter, tree depth, training, and testing ratios, could be processed and analyzed in RFC, which presents a tool that can produce the optimum result.
Therefore, this study considered RFC and optimized the parameters to produce the best tree to serve as a classification model. Several parameters for the criterion and splitter, such as gini or entropy and random or best, were considered, similar to the study of Ong et al. [23,24]. In addition, the tree depth was considered among four to seven and run through different training and testing ratios. A total of 6400 runs for 100 combinations each were employed.

3.3. Artificial Neural Network

ANN has a more complex calculation and algorithm compared to the RFC. ANN has been utilized among other studies considering natural disasters and human behavior. Moustra et al. [44] utilized this algorithm for earthquake prediction. Their study focused on historical datasets to analyze the geographic locations of earthquakes. Yariyan et al. [45] considered it for risk assessment mapping for the vulnerability of earthquakes. Oktarina et al. [46] studied earthquake casualties and damages in Indonesia with ANN. Their study presented higher pattern recognition with high accuracy upon utilizing ANN. Similarly, Ong et al. [24] considered RFC and ANN to predict factors affecting the acceptance of Bataan reopening, a decommissioned nuclear power plant in the Philippines. Due to their limitations, these studies presented how ANN could be considered instead of multivariate and traditional statistical tools.
ANN, according to Jamshidi et al. [47], Ong et al. [23,24], and Yuduang et al. [38], could be a classification tool best suited to analyze factors affecting human behavior. Most studies consider ANN before proceeding to other types of neural networks (e.g., Deep Learning). It was stated that if ANN’s accuracy and complexity power produces a low output, then deep learning may be considered. Nonetheless, ANN is considered sufficient when high predictive power is obtained. General neural networks are considered cutting-edge algorithms despite level. ANN was compared to other classification techniques, such as K-Nearest Neighbors and Naïve Bayes, which also have high predictive powers. However, Ong et al. [24] reiterated the applicability of ANN towards human behavior for relatively simple frameworks, similar to this study. Moreover, powerful algorithms should be applied if the study considers complex frameworks for analysis.
The input data for RFC, optimization for the parameters, several activation functions for the hidden layer, such as sigmoid, tanh, and relu, and the number of nodes for the input layer were considered for the generation of the ANN model utilizing Python 5.1. In addition, the activation functions for the output layer considered sigmoid and swish. Lastly, adam, SGD, and RMSProp were considered for the optimizers. In total, 16,200 runs were employed for ten combinations each at 150 epochs [48]. From the results, the best classification model considered parameters such as tanh for the hidden layer and sigmoid for the output and ran with adam with 50 nodes at the hidden layer. Further optimization for different training and testing ratios were conducted.

4. Results

4.1. Descriptive Statistics of the Items

The items utilized in this study were adopted from the study by Kurata et al. [2], as seen in the Appendix A of the manuscript. Presented in Table 1 are the descriptive statistical results of the items from the collected responses. It could be seen that average standard deviation and mean values were within the range of normality, ±1.96 using the Harman’s Single Factor Analysis [20]. In addition, Cronbach’s alpha results presented acceptable values, greater than 0.70, which indicated that the collected data could be used to assess factors affecting volcanic eruption preparedness among Filipinos.

4.2. Random Forest Classifier

Figure 3 represents the best tree with RFC among all results tested. From the figure, it could be seen that the parent node indicated media (X1) as the factor dictating preparedness for a volcanic eruption, with a value less than, or equal to, 0.386. Satisfying this would consider perceived severity (X0) with values less than, or equal to, 2.13. Satisfying this would consider X0, X1, and perceived vulnerability (X2), leading to high preparedness for a volcanic eruption. However, if X0 was not satisfied, it would consider X1 and X2, leading to high preparedness for a volcanic eruption.
On the other hand, if the parent node would not be satisfied, it would consider X0 with values less than, or equal to, 0.123. Satisfying this condition would consider X1 and X2 leading to high preparedness for a volcanic eruption. If this was not satisfied, it would consider X1 and intention (X3) with values less than, or equal to, −0.074 leading to very high preparedness. If the child node was not satisfied, it would consider X2, X3, and X2 leading to high preparedness, indicating that X0, X1, and X3 were the highly significant factors affecting preparedness for a volcanic eruption. X2, on the other hand, was considered a significant factor that would only highly affect preparedness.
Table 2 represents the summary of accuracy results for the optimization. At depth 5, the best tree with RFC was produced. A 93% accuracy with a 0.00 standard deviation was seen as the highest among the results. Utilizing ANOVA, no significant difference was seen; thus, the highest accuracy with the lowest standard deviation was considered for the model. Gini and Best as the criteria and splitter at 80:20 training and testing ratio produced the best tree.
Figure 4 represents the scatter plot for the RFC accuracy results among Gini Index Criterion. It could be seen that the scatter plot peaked at the highest average accuracy of 93%.

4.3. Artificial Neural Network

Table 3 presents the summary of ANN after the final optimization run. At 200 epochs, the final results from the 80:20 training and testing ratio are presented. According to the study of Ong et al. [23,24], the average testing sequence dictates the significance ranking among the latent variables considered. It could be deduced that media (M) presented as the contributing factor affecting preparedness for a volcanic eruption, followed by perceived severity (PS), perceived vulnerability (PV), and intention (IN). Other significant factors identified were perceived risk proximity (PR), subjective norm (SN), hazard knowledge (HK), perceived behavioral control (PBC), and attitude (AT). The threshold was 60%, and anything lower was not considered significant [34].
Figure 5 represents the scatter plot for the ANN average testing accuracy. It could be seen that the sequence of the results presented Media as the highest latent variable, followed by Perceived Severity, Perceived Vulnerability, Intentions, Perceived Risk Proximity, Subjective Norm, Hazard Knowledge, Perceived Behavioral Control, and Attitude as the lowest.
The score of importance was also computed to verify the findings of ANN. Similar results of significant factor ranking were seen, as presented in Table 4. In addition, the training and validation loss rate was also assessed, as presented in Figure 6. Following the suggestion of Lara et al. [49], the figure indicated no overfitting when the training and validation loss rates were relatively close with an area coinciding. If the rates were far above (below) each other, it indicated overfitting (underfitting).
With 50 nodes in the hidden layer, using Tanh and Sigmoid as activation functions for the hidden and output layers, and considering adam as the optimizer, produced an accuracy of 98.86%. Presented in Figure 7 is the optimum ANN model produced from the parameters.

4.4. Evaluation of Accuracy

To further evaluate the accuracy of the method used, ANN and RFC, different analyses, such as the Taylor Diagram, Violin Plot, and Box Plot, were conducted utilizing Python 5.1 using the seaborn package. Gholami et al. [50] explained how the Taylor Diagram assesses the performance of the model based on accuracy, showing the standard deviation and correlation. From their study, a threshold of Root Mean Square Error less than 20% was considered acceptable, while correlation greater than 90% was deemed significant. In this study, Figure 8 represents the Taylor Diagram showing the accuracy rate of RFC and ANN accuracies for different latent variables tested for their effect on preparedness of people. It could be deduced that RFC accuracy was within the threshold of highly significant results, together with latent variables M, PS, PV, IN, PR, SN, HK, and PBC. On the other hand, AT showed parameters outside the threshold, which indicated its low significance with regard to preparedness. In line with the results of individual analysis, AT was not part of the highly significant factors in RFC and least so in ANN. Thus, consistency for the accuracy rate was seen.
In addition, the Violin Plot to analyze the percentile ranking of the accuracies was conducted, together with the Box Plot, as seen in Figure 9 and Figure 10. Figure 9 shows that that the accuracy mean for the RFC algorithm was at 90%. From the results, the highest accuracy obtained was 93%, which showed that it was within the higher level of interquartile range. Similarly, the ANN analysis showed a 94% accuracy mean. Until 83%, the accuracies would be considered in the interquartile range and anything below would be set for the lower adjacent values. In line with the results of the study, M, PS, PV, IN, PR, SN, HK, and PBC were among the interquartile range, while AT showed accuracy below. This showed how likely the AT latent variable was less significant compared to the others, which was similar to the findings of ANN and the Taylor Diagram. Similarly, the Box Plot in Figure 10 presented consistent results. Thus, it could be deduced that there was a uniform distribution seen from the Violin Plot results [51].

5. Discussion

The evident increase in natural disasters, such as volcanic eruptions, has been widely prominent [2]. With people experiencing casualties brought by volcanic eruptions, this study considered an MLA ensemble with ANN and RFC to predict factors affecting preparedness for a volcanic eruption. The RFC produced a classification model with 93% accuracy resulting in media (M) being the most significant factor affecting preparedness for a volcanic eruption. Following which was perceived severity (PS), perceived vulnerability (PV), and intention (IN), which were all highly significant factors. ANN presented the same ranking of results with 98.86% accuracy, indicating that other factors, such as perceived risk proximity (PR), subjective norm (SN), hazard knowledge (HK), perceived behavioral control (PBC), and attitude (AT), were also considered significant.
Media (M) was seen to be a significant factor (100%), with its indicators presenting that people use media to quickly understand and obtain knowledge and correct information regarding threats of volcanic eruption=. It could also be deduced that people know which media platforms promptly report on threats brought by natural disasters, such as volcanic eruptions. Ong et al. [20] indicated that media is one of the most significant factors that indirectly affect the intention to prepare among communities when a threat of a natural disaster is present. Kurata et al. [21] support the findings by indicating how mass media and the development of technology have helped spread information about natural calamities relatively faster. The presence and availability of media have therefore been widely utilized to gain information for the community regarding disasters or calamities and to create positive social information dissemination [52]. Media is the highest contributing factor, as seen from the resultsa, since volcanic eruptions are among the most unpredictable natural disaster which could be active at any time. Use of media, allows people to monitor how active a volcano is, leading to more time for mitigation and preparation.
Relative to the current event of the Taal volcano being active and Mt. Bulusan in the Philippines, the media, through television, radio, and even social media, were able to indicate the level of activity, even giving precautions about activity before it began. Thus, the community relies heavily on media for them to have prepared for a volcanic eruption. However, studies like that of Ramakrishnan et al. [53] highlighted the applicability of Media in general. Of course, those without access to it are unable to utilize it, especially those living in underserved communities. It was also highlighted that the individual intends to know the significance through media access, similar to Armstrong et al.’s findings [54]. If people have difficulty in access, then Media would be considered an insignificant latent variable [53,54].
Second, PS was seen to have a highly significant effect (98.6%) on preparedness, much higher than the RFC result. People are highly aware of the effects of volcanic eruption on livelihood, fatalities, severity of aftermath, and the fact that volcanic eruptions can cause economic crisis. In line with this, SN was seen to have a highly significant effect on preparedness (94.3%). Andreastuti et al. [55] explained how perceptions would still be the critical factor affecting actions and attitude towards preparation and mitigation. When people know how an eruption will affect people who are important to them, their health, and their lives, people are more inclined to know how to deal with the cause [28]. Cahigas et al. [56] also indicated that crisis management would be one of the highest factors affecting an individual’s behavior when a threat to health is present. Barclay et al. [57] highlighted that the community would focus on wellbeing and livelihood when natural disaster threats are present and would be highly incentivized by their perception of severity during and after a natural disaster like a volcanic eruption. Thus, this supports how PS presents a high indicator of preparedness for a volcanic eruption. However, if people perceive the natural disaster to have relatively low severity, then the preparedness and intention are low [57].
In line with PS, PV significantly affected preparedness for volcanic eruption (98.5%). With the recent events of the Taal volcano eruption in the country, people experienced how volcanic eruptions would affect a wide range of areas. The Taal was within a 100-km radius of the country’s capital, which felt the movement and ash fall despite the minimal eruption. People indicated they were highly vulnerable if the Taal volcano fully erupted. Mt. Bulusan, on the other hand, is within a 570-km radius of the capital, but the community was also seen to be vigilant. This indicates how the experience of volcanic eruption promotes PV. It was indicated by Weichselgartner and Pigeon [25] that people are more inclined to prepare when they have knowledge, understanding, and experience of natural disasters such as volcanic eruptions. On another note, Warsini et al. [58] presented that more people would have camaraderie and mitigation plans due to the knowledge acquired from experience of volcanic eruptions due to their PV.
Fourth, IN was seen to be a significant factor affecting preparedness (97.2%). People willingly followed evacuation plans and safety measures set by the government if volcanic eruptions were bound to happen. The primary factors, such as PS and PV, would support this finding. Since people have the knowledge and experience, they would have high PS and PV, leading them to follow safety protocols. Aside from their mitigation plans, the evacuation from homes would be challenging, especially with 79% answering that they own their homes. Hershkovich et al. [59] and Kurata et al. [2] highlighted that with high PS and PV, people’s IN would be more aligned with the government plans. It was also indicated that the government and other authorities are highly influential, due to their practical actions with evacuation and mitigation plans. Thus, the community would have high IN for preparation during a volcanic eruption.
Fifth, PR (96.8%) and HK (89.8%) were significant factors. It could be deduced that people understand and know about volcanic eruptions. In addition, people are familiar with how close or far they are to the proximity of volcanoes, and their preparedness is highly influenced when they are much closer. Martinez-Villegas et al. [60] showed the direct relationship between evacuation response during a volcanic eruption. Gaillard [61] showed that people living near active volcanoes are more likely to respond to preparation and mitigation due to economic factors. Baxter et al. [62] explored mitigation and preparation plans for volcanic eruptions due to limited experience and infrequent events. This relationship showed that intention and preparedness are insignificant if risks and hazard knowledge are low [62]. However, the past years in the Philippines have proposed high eruption rates due to the constant threat posed by the Taal volcano and now Mt. Bulusan. Perry and Lindell [63] showed how the effect of natural disasters near a community would enhance their preparedness. Thus, it presents a higher threat among people in proximity, leading to higher preparation.
It was seen that PBC (83.2%) and AT (79.9%) were significant factors affecting preparation for a volcanic eruption. With regard to t behavioral aspect of an individual’s control over eruption hazards, how they can be avoided, having preventive measures, and knowledge of appropriate action presented significant indicators. A positive AT regarding security, confidence, and concern regarding safety in terms of health and life was evident. It was explained by Ong et al. [20] that TPB factors such as PBC and AT have a highly significant direct effect on the intention to prepare for mitigation of a natural disaster. Similar results were seen from the study by Morganstein and Ursano [64] and Abella et al. [65], wherein they highlighted and explained that awareness of a natural disaster would lead to an increase in response due to gaining knowledge of disaster preparation. Kusumastuti et al. [66] also highlighted how an individual’s attitude and experience affected control over behavior and would develop and enhance alertness and preparation. Concerning the previous discussion, knowledge, PS, and PV would lead to a positive AT and PBC among individuals if they had experience and understanding of the natural disaster, its effects, and consequential aftermath [20]. Therefore, the more likely an individual is to be affected by a natural disaster, the more they would be inclined to mitigation and preparation.
It could be deduced that the government’s protocol, policies, plans, and preemptive measures are highly effective in preparing for a volcanic eruption. In addition, the threat posed by the natural disaster, perceived severity and vulnerability, and knowledge would lead to remarkably high preparedness, among other factors. If people have experienced the effects of volcanic eruptions, they would be more inclined to prepare and mitigate. The effects would lead to a more positive behavioral control and positive attitude. Thus, the government may highlight the adverse effects of volcanic eruptions to help people adapt and prepare for volcanic eruptions.

5.1. Theoretical and Practical Implications

The results presented the high accuracy of classification models created from machine learning algorithm ensembles. Despite the non-linear relationship presented in the framework, a highly significant relationship was evident among all factors. In addition, in contrast to the findings of Kurata et al. [2], it was seen that several relationships prompted insignificant results, which signified that the claim by Fan et al. [12] and Woody [13] should be highly considered upon the utilization of structural equation modeling (SEM). In addition, the farther the dependent variable (e.g., media), the lower the relationship value. It was presented in Kurata et al. [2,21] that media was the fourth highest significant factor; however, this study proved it to be the most significant among others. This shows that the arrangement of latent variables affected their results with SEM. Thus, utilizing an MLA ensemble could provide better output than the multivariate analysis promoted by SEM. This study contributed to the MLA ensemble usage, specifically in the field of natural disaster and human behavior related studies. It was indicated by Hanel et al. [67] that there is a difference in overall behavior in every country. In addition, different intentions and practices would be evident with different types of natural disasters, due to perceived severity, perceived vulnerability, and knowledge.
The findings suggest that PS and PV could be critical factors for preparing for a volcanic eruption. It could be posited that media may be highly utilized to spread awareness and information regarding natural disasters. Therefore, the government may utilize these attributes to enhance their mitigation and preparation plans. In addition, the relative behavioral aspects may also be considered when people have knowledge and experience. Thus, the government may consider a theoretical approach to address the aftermath of a volcanic eruption. The findings could also be classified among the community for them to visualize and understand what volcanic eruptions may cause during and after the said event. Since the Philippines is currently monitoring volcanic activities through the Philippine Institute of Volcanology and Seismology (PHIVOLCS), the government may consider the reports and information obtained for mitigation plans. Preparing before the disaster would help save livelihoods, reduce economic crisis, and save lives.

5.2. Limitations

Despite the relative findings, several limitations are still considered. First, this study considered a self-administered online questionnaire. Other factors were, therefore, not considered since the study had a theoretical-framework-based approach. It is suggested to conduct interviews among citizens, especially those living in proximity to volcanoes to create qualitative results. This way, researchers could dive deeper into the findings and it would help the government create action and mitigation plans. Second, it is suggested to consider specific respondents, such as those with and without access to media, the less educated, and rural versus urban areas. The results may present comparative findings of differences in educational attainment, hazard knowledge, and access to information. Third, other MLA could be utilized, such as clustering techniques, since this study created a classification model. The clustering may help segregate the demographic factors of the respondents. Lastly, it is suggested to test higher calculation complexity algorithms to promote the utility of an MLA ensemble regarding natural disasters and human behavior. Other classification tools may be considered, such as clustering, other types of neural networks, such as K-nearest neighbors and Naïve Bayes, and the like to compare and contrast the results, create a benchmark, and provide algorithms applicable in natural disaster-related and human behavior studies.

6. Conclusions

The increasing number of volcanic eruptions in the Philippines has been evident, with the Taal volcano erupting in early 2020 and Mt. Bulusan in 2022. Moreover, various countries have reported increasing deaths and damage from natural disasters (e.g., volcanic eruptions) over the past few years. However, the need to assess preparedness for volcanic eruption has been underexplored. This study evaluated the factors that could influence preparedness for volcanic eruptions. Utilizing a machine learning ensemble, a high accuracy rate was seen with RFC (93%) and ANN (98.86%).
The results indicated that the media, being the main source of information, had the highest effect on preparedness, followed by PS, PV and IN as contributing factors to high preparedness for a volcanic eruption. The effect on lives, economy and livelihood led to increased community preparedness. The proximity was seen to be a significant factor, followed by SN, and HK, which indicated how knowledge and experience, the effect on the individual and the people important to them caused a significant relationship. Lastly, PBC and AT showed how people with control over their behavior would have positive attitudes to preparation for volcanic eruption, following the government’s mitigation and preparation plan.
It was seen that people would likely follow protocols and plans presented by the government when natural disasters such as volcanic eruptions occur. Thus, the government may capitalize on this and present a volcanic eruption’s theoretical aftermath and effects to enhance people’s activeness and support towards preparedness. With its highly accurate results, the methodology, framework, and results of this study may be extended to evaluate other natural disasters. In addition, this could also be applied to other related studies across the world. With the threat of climate change, the need to reevaluate behavior and mitigation plans is needed. Thus, this study contributes to the safety and livelihood of people during natural disasters, such as volcanic eruptions.

Author Contributions

Conceptualization, J.D.G., A.A.N.P.R., A.K.S.O. and Y.T.P.; methodology, J.D.G., A.A.N.P.R., V.L.M.S. and A.K.S.O.; software, A.K.S.O., Y.T.P. and V.L.M.S.; validation, J.D.G. and A.A.N.P.R.; formal analysis, A.K.S.O. and Y.T.P.; investigation, J.D.G. and A.A.N.P.R.; resources; A.K.S.O. and Y.T.P.; data curation, Y.T.P.; writing-original draft preparation, J.D.G.; A.A.N.P.R.; writing-review and editing, J.D.G., A.A.N.P.R. and A.K.S.O.; visualization, A.K.S.O. and Y.T.P.; supervision, A.K.S.O. and A.A.N.P.R.; project administration, Y.T.P.; and funding acquisition, Y.T.P. All authors have read and agreed to the published version of the manuscript.

Funding

This research was funded by Mapua University Directed Research for Innovation and Value Enhancement (DRIVE).

Institutional Review Board Statement

This study was approved by Mapua University Research Ethics Committees (FM-RC-22-17).

Informed Consent Statement

Informed consent was obtained from all subjects involved in this study (FM-RC-21-54).

Data Availability Statement

The data presented in this study are available on request from the corresponding author.

Acknowledgments

The authors would like to thank all the respondents who answered our online questionnaire. We would also like to thank our friends for their contributions in the distribution of the questionnaire.

Conflicts of Interest

The authors declare no conflict of interest.

Appendix A

Table
Table A1. Questionnaire (with permission from Kurata et al. [2]. 2022, Elsevier).
Table A1. Questionnaire (with permission from Kurata et al. [2]. 2022, Elsevier).
ConstructItemsMeasures
MediaMP1I think there are a variety of sources for media information about volcanic activities.
MP2I believe that the information among media is easily shared.
MP3I believe that social media contributes to the quick spreading of information regarding volcanic eruptions.
MP4I know how to distinguish irrelevant information (fake news) from social media platforms.
Hazard KnowledgeHK1I am familiar with the disaster sirens and warning signals.
HK2I believe I should wear masks to protect me from inhaling the ashes during volcanic eruption.
HK3I know that I should limit myself from going outdoors to reduce my ash exposure.
HK4I know that I should wear protective clothing during volcano eruption if I go outdoors.
HK5I know that there is poor air quality before, during, and after a volcanic eruption.
Perceived Risk ProximityGP1I think my location is within the danger zone from a volcano.
GP2I am familiar with the nearest evacuation facility I can go to if a volcano erupts.
GP3I am aware of the risks I have from volcaniv eruption based on my location.
Perceived SeverityPS1I believe that a Taal volcano eruption is severe.
PS2I believe that a Taal volcano eruption may lead to deaths among people.
PS3I believe that a Taal volcano eruption is much more severe than other volcanic eruptions.
PS4I find that a Taal volcano eruption may affect my livelihood.
PS5I think that it will cpst me much to rebuild my resources affected by a Taal volcano eruption.
Perceived VulnerabilityPV1I think my community is vulnerable to experience the effects of Taal volcanic eruption.
PV2I am likely to experience the effects of a Taal volcano eruption based on my experience.
PV3I have an experience of being vulnerable to volcano eruptions.
PV4I know I am more vulnerable to severe effects of volcanic eruption if I have breathing problems.
Perceived Behavioral ControlPBC1I am in control of the situation in protecting myself from eruption hazards.
PBC2I think it is easy to implement preventive measures in my vicinity.
PBC3I know that I can avoid experiencing the effects of a volcanic eruption.
PBC4I think I have sufficient knowledge in responding to effects of volcanic eruption.
Subjective NormsSN1Most people in my community follow the preventive measures given by the local government unit.
SN2People in my community receivedaid from the local government unit.
SN3People living in my community go to the assigned evacuation centers before the volcano erupts.
SN4Most people in my community observe safety measures when the volcano shows possibility of eruption.
SN5People in my community still work for their living even though the government raises the alert levels for volcanic eruption.
Attitude toward the BehaviorATB1I am stressed in a volcanic eruption.
ATB2I am scared if my family will be affected by a volcanic eruption.
ATB3I feel insecure when my community does not prepare for volcanic eruption.
ATB4I am concerned when the government raises the alert level of the Taal volcano.
ATB5I am confident that I know how to respond when a volcano erupts.
Intention to EvacuateIF1I am willing to adhere to the authorities’ instructions if they tell me to evacuate.
IF2I am willing to leave anything behind to put myself in the safest place in the fastest way possible.
IF3My family is willing to leave anything behind to be in the safest place in the fastest way possible.
IF4I am willing to stay in the evacuation area with other people until it is safe to go back in my home.
PreparednessPP1I believe preemptive measures by the authorities for disaster response are effective.
PP2I believe that the emergency warning awareness will keep me safe during volcanic eruption.
PP3I believe that being updated by mass media will keep me safe during volcanic eruption.
PP4I think being connected to any friend or family member who does not live near the volcano is necessary in any case of a life-threatening circumstance.
PP5I think that it is essential to evacuate early if I am living in a community with the greatest risk.

References

  1. Coppola, D.; Laiolo, M.; Cigolini, C.; Massimetti, F.; Delle Donne, D.; Ripepe, M.; Arias, H.; Barsotti, S.; Parra, C.B.; Centeno, R.G.; et al. Thermal Remote Sensing for Global Volcano Monitoring: Experiences from the mirova system. Front. Earth Sci. 2020, 7, 362. [Google Scholar] [CrossRef]
  2. Kurata, Y.B.; Prasetyo, Y.T.; Ong, A.K.; Nadlifatin, R.; Persada, S.F.; Chuenyindee, T.; Cahigas, M.M. Determining factors affecting preparedness beliefs among Filipinos on Taal Volcano eruption in Luzon, Philippines. Int. J. Disaster Risk Reduct. 2022, 76, 103035. [Google Scholar] [CrossRef]
  3. Brown, S.K.; Jenkins, S.F.; Sparks, R.S.; Odbert, H.; Auker, M.R. Volcanic fatalities database: Analysis of volcanic threat with distance and victim classification. J. Appl. Volcanol. 2017, 6, 15. [Google Scholar] [CrossRef]
  4. Niroa, J.J.; Nakamura, N. Volcanic disaster risk reduction in indigenous communities on Tanna Island, Vanuatu. Int. J. Disaster Risk Reduct. 2022, 74, 102937. [Google Scholar] [CrossRef]
  5. Thouret, J.-C.; Wavelet, E.; Taillandier, M.; Tjahjono, B.; Jenkins, S.F.; Azzaoui, N.; Santoni, O. Defining population socio-economic characteristics, hazard knowledge and risk perception: The adaptive capacity to persistent volcanic threats from Semeru, Indonesia. Int. J. Disaster Risk Reduct. 2022, 77, 103064. [Google Scholar] [CrossRef]
  6. Reyes-Hardy, M.-P.; Aguilera Barraza, F.; Sepúlveda Birke, J.P.; Esquivel Cáceres, A.; Inostroza Pizarro, M. GIS-based volcanic hazards, vulnerability and risks assessment of the Guallatiri Volcano, Arica y Parinacota Region, Chile. J. S. Am. Earth Sci. 2021, 109, 103262. [Google Scholar] [CrossRef]
  7. Barone, G.; De Giudici, G.; Gimeno, D.; Lanzafame, G.; Podda, F.; Cannas, C.; Giuffrida, A.; Barchitta, M.; Agodi, A.; Mazzoleni, P. Surface reactivity of Etna Volcanic Ash and evaluation of Health Risks. Sci. Total Environ. 2021, 761, 143248. [Google Scholar] [CrossRef]
  8. Michellier, C.; Kervyn, M.; Barette, F.; Muhindo Syavulisembo, A.; Kimanuka, C.; Kulimushi Mataboro, S.; Hage, F.; Wolff, E.; Kervyn, F. Evaluating population vulnerability to volcanic risk in a data scarcity context: The case of Goma City, Virunga Volcanic Province (DRCongo). Int. J. Disaster Risk Reduct. 2020, 45, 101460. [Google Scholar] [CrossRef]
  9. Dogar, M.M.; Sato, T. Regional climate response of middle eastern, African, and South Asian monsoon regions to explosive volcanism and Enso forcing. J. Geophys. Res. Atmos. 2019, 124, 7580–7598. [Google Scholar] [CrossRef]
  10. Dogar, M.M.; Sato, T. Analysis of climate trends and leading modes of climate variability for Mena Region. J. Geophys. Res. Atmos. 2018, 123, 13074–13091. [Google Scholar] [CrossRef]
  11. Dogar, M.M.; Stenchikov, G.; Osipov, S.; Wyman, B.; Zhao, M. Sensitivity of the regional climate in the Middle East and North Africa to volcanic perturbations. J. Geophys. Res. Atmos. 2017, 122, 7922–7948. [Google Scholar] [CrossRef]
  12. Fan, Y.; Chen, J.; Shirkey, G.; John, R.; Wu, S.R.; Park, H.; Shao, C. Applications of structural equation modeling (SEM) in Ecological Studies: An updated review. Ecol. Processes 2016, 5, 19. [Google Scholar] [CrossRef]
  13. Woody, E. An SEM perspective on evaluating mediation: What every clinical researcher needs to know. J. Exp. Psychopathol. 2011, 2, 210–251. [Google Scholar] [CrossRef]
  14. Duarte, P.; Pinho, J.C. A mixed methods UTAUT2-based approach to assess mobile health adoption. J. Bus. Res. 2019, 102, 140–150. [Google Scholar] [CrossRef]
  15. PHIVOLCS Taal Volcano Bulletin 9 April 2022 08:00 am. Available online: https://www.phivolcs.dost.gov.ph/index.php/volcano-hazard/volcano-bulletin2/taal-volcano/14430-taal-volcano-bulletin-9-april-2022-08-00-am#:~:text=There%20has%20been%20no%20recorded,dropped%20on%203%20April%202022 (accessed on 31 May 2022).
  16. Mangosing, M.A.M.-M.F. Bulusan Eruption Rains Ash, Forces Evacuation. Available online: https://newsinfo.inquirer.net/1606790/bulusan-eruption-rains-ash-forces-evacuation (accessed on 31 May 2022).
  17. Manila, U.S.E. Natural Disaster Alert–Mount Bulusan at Alert Level 1, June 6, 2022. Available online: https://ph.usembassy.gov/natural-disaster-alert-mount-bulusan-at-alert-level-1/ (accessed on 25 June 2022).
  18. The World Data Active Volcanoes and Eruptions in the Philippines. Available online: https://www.worlddata.info/asia/philippines/volcanos.php (accessed on 25 February 2022).
  19. Gumasing, M.J.; Prasetyo, Y.T.; Ong, A.K.; Nadlifatin, R. Determination of factors affecting the response efficacy of Filipinos under Typhoon Conson 2021 (jolina): An extended protection motivation theory approach. Int. J. Disaster Risk Reduct. 2022, 70, 102759. [Google Scholar] [CrossRef]
  20. Ong, A.K.; Prasetyo, Y.T.; Lagura, F.C.; Ramos, R.N.; Sigua, K.M.; Villas, J.A.; Young, M.N.; Diaz, J.F.; Persada, S.F.; Redi, A.A. Factors affecting intention to prepare for mitigation of “The big one” earthquake in the Philippines: Integrating protection motivation theory and extended theory of planned behavior. Int. J. Disaster Risk Reduct. 2021, 63, 102467. [Google Scholar] [CrossRef]
  21. Kurata, Y.B.; Prasetyo, Y.T.; Ong, A.K.; Nadlifatin, R.; Chuenyindee, T. Factors affecting perceived effectiveness of typhoon vamco (Ulysses) flood disaster response among Filipinos in Luzon, Philippines: An integration of protection motivation theory and extended theory of planned behavior. Int. J. Disaster Risk Reduct. 2022, 67, 102670. [Google Scholar] [CrossRef]
  22. Prasetyo, Y.T.; Castillo, A.M.; Salonga, L.J.; Sia, J.A.; Seneta, J.A. Factors affecting perceived effectiveness of COVID-19 prevention measures among Filipinos during enhanced community quarantine in Luzon, Philippines: Integrating Protection Motivation Theory and extended theory of planned behavior. Int. J. Infect. Dis. 2020, 99, 312–323. [Google Scholar] [CrossRef]
  23. Ong, A.K.; Chuenyindee, T.; Prasetyo, Y.T.; Nadlifatin, R.; Persada, S.F.; Gumasing, M.J.; German, J.D.; Robas, K.P.; Young, M.N.; Sittiwatethanasiri, T. Utilization of random forest and deep learning neural network for predicting factors affecting perceived usability of a COVID-19 contact tracing mobile application in Thailand “Thaichana. ” Int. J. Environ. Res. Public Health 2022, 19, 6111. [Google Scholar] [CrossRef] [PubMed]
  24. Ong, A.K.; Prasetyo, Y.T.; Velasco, K.E.; Abad, E.D.; Buencille, A.L.; Estorninos, E.M.; Cahigas, M.M.; Chuenyindee, T.; Persada, S.F.; Nadlifatin, R.; et al. Utilization of random forest classifier and artificial neural network for predicting the acceptance of reopening decommissioned nuclear power plant. Ann. Nucl. Energy 2022, 175, 109188. [Google Scholar] [CrossRef]
  25. Weichselgartner, J.; Pigeon, P. The role of knowledge in disaster risk reduction. Int. J. Disaster Risk Sci. 2015, 6, 107–116. [Google Scholar] [CrossRef]
  26. Phengsuwan, J.; Shah, T.; Thekkummal, N.B.; Wen, Z.; Sun, R.; Pullarkatt, D.; Thirugnanam, H.; Ramesh, M.V.; Morgan, G.; James, P.; et al. Use of social media data in Disaster Management: A survey. Future Internet 2021, 13, 46. [Google Scholar] [CrossRef]
  27. Dube, E.; Munsaka, E. The contribution of indigenous knowledge to disaster risk reduction activities in Zimbabwe: A big call to practitioners. Jàmbá J. Disaster Risk Stud. 2018, 10, 493. [Google Scholar] [CrossRef] [PubMed]
  28. Chuenyindee, T.; Ong, A.K.; Prasetyo, Y.T.; Persada, S.F.; Nadlifatin, R.; Sittiwatethanasiri, T. Factors affecting the perceived usability of the COVID-19 contact-tracing application “Thai chana” during the early COVID-19 omicron period. Int. J. Environ. Res. Public Health 2022, 19, 4383. [Google Scholar] [CrossRef]
  29. Arias, J.P.; Bronfman, N.C.; Cisternas, P.C.; Repetto, P.B. Hazard proximity and risk perception of tsunamis in coastal cities: Are people able to identify their risk? PLoS ONE 2017, 12, e0186455. [Google Scholar] [CrossRef] [PubMed]
  30. Yagoub, M.M.; Al Yammahi, A.A. Spatial distribution of natural hazards and their proximity to Heritage Sites: Case of the United Arab Emirates. Int. J. Disaster Risk Reduct. 2022, 71, 102827. [Google Scholar] [CrossRef]
  31. Rana, I.A.; Jamshed, A.; Younas, Z.I.; Bhatti, S.S. Characterizing flood risk perception in urban communities of Pakistan. Int. J. Disaster Risk Reduct. 2020, 46, 101624. [Google Scholar] [CrossRef]
  32. German, J.D.; Redi, A.A.; Prasetyo, Y.T.; Persada, S.F.; Ong, A.K.; Young, M.N.; Nadlifatin, R. Choosing a package carrier during COVID-19 pandemic: An integration of pro-environmental planned behavior (PEPB) theory and Service Quality (SERVQUAL). J. Clean. Prod. 2022, 346, 131123. [Google Scholar] [CrossRef] [PubMed]
  33. Vinnell, L.J.; Milfont, T.L.; McClure, J. Why do people prepare for natural hazards? developing and testing a theory of planned behaviour approach. Curr. Res. Ecol. Soc. Psychol. 2021, 2, 100011. [Google Scholar] [CrossRef]
  34. Najafi, M.; Ardalan, A.; Akbarisari, A.; Noorbala, A.A.; Jabbari, H. Demographic determinants of Disaster Preparedness Behaviors amongst Tehran inhabitants, Iran. PLoS Curr. 2015, 7, 1–15. [Google Scholar] [CrossRef]
  35. Bronfman, N.C.; Cisternas, P.C.; Repetto, P.B.; Castañeda, J.V. Natural disaster preparedness in a multi-hazard environment: Characterizing the sociodemographic profile of those better (worse) prepared. PLoS ONE 2019, 14, e0214249. [Google Scholar] [CrossRef] [PubMed]
  36. Bourque, L.B.; Regan, R.; Kelley, M.M.; Wood, M.M.; Kano, M.; Mileti, D.S. An examination of the effect of perceived risk on preparedness behavior. Environ. Behav. 2012, 45, 615–649. [Google Scholar] [CrossRef]
  37. Heller, K.; Alexander, D.B.; Gatz, M.; Knight, B.G.; Rose, T. Social and personal factors as predictors of earthquake preparation: The role of support provision, network discussion, negative affect, age, and EDUCATION1. J. Appl. Soc. Psychol. 2005, 35, 399–422. [Google Scholar] [CrossRef]
  38. Yuduang, N.; Ong, A.K.; Vista, N.B.; Prasetyo, Y.T.; Nadlifatin, R.; Persada, S.F.; Gumasing, M.J.; German, J.D.; Robas, K.P.; Chuenyindee, T.; et al. Utilizing structural equation modeling–artificial neural network hybrid approach in determining factors affecting perceived usability of mobile mental health application in the Philippines. Int. J. Environ. Res. Public Health 2022, 19, 6732. [Google Scholar] [CrossRef] [PubMed]
  39. Podsakoff, P.M.; MacKenzie, S.B.; Lee, J.-Y.; Podsakoff, N.P. Common method biases in behavioral research: A critical review of the literature and recommended remedies. J. Appl. Psychol. 2003, 88, 879–903. [Google Scholar] [CrossRef]
  40. Kim, Y.; Hardisty, R.; Torres, E.; Marfurt, K.J. Seismic-facies classification using random forest algorithm. In SEG Technical Program Expanded Abstracts 2018; SEG Library: Anaheim, CA, USA, 2018. [Google Scholar]
  41. Snehil; Goel, R. Flood damage analysis using machine learning techniques. Procedia Comput. Sci. 2020, 173, 78–85. [Google Scholar] [CrossRef]
  42. Chen, J.; Li, Q.; Wang, H.; Deng, M. A machine learning ensemble approach based on Random Forest and radial basis function neural network for risk evaluation of Regional Flood Disaster: A case study of the yangtze river delta, China. Int. J. Environ. Res. Public Health 2019, 17, 49. [Google Scholar] [CrossRef]
  43. Yang, W.; Zhou, S. Using decision tree analysis to identify the determinants of residents’ CO2 emissions from different types of trips: A case study of guangzhou, China. J. Clean. Prod. 2020, 277, 124071. [Google Scholar] [CrossRef]
  44. Moustra, M.; Avraamides, M.; Christodoulou, C. Artificial Neural Networks for earthquake prediction using time series magnitude data or seismic electric signals. Expert Syst. Appl. 2011, 38, 15032–15039. [Google Scholar] [CrossRef]
  45. Yariyan, P.; Zabihi, H.; Wolf, I.D.; Karami, M.; Amiriyan, S. Earthquake risk assessment using an integrated fuzzy analytic hierarchy process with artificial neural networks based on GIS: A case study of sanandaj in Iran. Int. J. Disaster Risk Reduct. 2020, 50, 101705. [Google Scholar] [CrossRef]
  46. Oktarina, R.; Bahagia, S.N.; Diawati, L.; Pribadi, K.S. Artificial Neural Network for predicting earthquake casualties and damages in Indonesia. IOP Conf. Ser. Earth Environ. Sci. 2020, 426, 012156. [Google Scholar] [CrossRef]
  47. Jamshidi, M.B.; Lalbakhsh, A.; Talla, J.; Peroutka, Z.; Roshani, S.; Matousek, V.; Roshani, S.; Mirmozafari, M.; Malek, Z.; La Spada, L.; et al. Deep learning techniques and COVID-19 drug discovery: Fundamentals, state-of-the-art and Future Directions. Stud. Syst. Decis. Control 2021, 348, 9–31. [Google Scholar] [CrossRef]
  48. Satwik, P.M.; Sundram, M. An integrated approach for weather forecasting and disaster prediction using Deep Learning Architecture based on memory augmented neural network’s (Mann’s). Mater. Today Proc. 2021; in press. [Google Scholar] [CrossRef]
  49. Lara, F.; Lara-Cueva, R.; Larco, J.C.; Carrera, E.V.; León, R. A deep learning approach for automatic recognition of seismo-volcanic events at the Cotopaxi Volcano. J. Volcanol. Geotherm. Res. 2021, 409, 107142. [Google Scholar] [CrossRef]
  50. Gholami, H.; Mohamadifar, A.; Sorooshian, A.; Jansen, J.D. Machine-learning algorithms for predicting land susceptibility to dust emissions: The case of the Jazmurian Basin, Iran. Atmos. Pollut. Res. 2020, 11, 1303–1315. [Google Scholar] [CrossRef]
  51. Thrun, M.C.; Gehlert, T.; Ultsch, A. Analyzing the fine structure of distributions. PLoS ONE 2020, 15, e0238835. [Google Scholar] [CrossRef] [PubMed]
  52. Sennert, S.S.; Klemetti, E.W.; Bird, D.K. Role of social media and networking in volcanic crises and Communication. Adv. Volcanol. 2015, 1, 733–743. [Google Scholar] [CrossRef]
  53. Ramakrishnan, T.; Ngamassi, L.; Rahman, S. Examining the factors that influence the use of social media for disaster management by underserved communities. Int. J. Disaster Risk Sci. 2022, 13, 52–65. [Google Scholar] [CrossRef]
  54. Armstrong, C.L.; Cain, J.A.; Hou, J. Ready for disaster: Information seeking, media influence, and disaster preparation for severe weather outbreaks. Atl. J. Commun. 2020, 29, 121–135. [Google Scholar] [CrossRef]
  55. Andreastuti, S.; Paripurno, E.T.; Gunawan, H.; Budianto, A.; Syahbana, D.; Pallister, J. Character of community response to volcanic crises at Sinabung and Kelud Volcanoes. J. Volcanol. Geotherm. Res. 2019, 382, 298–310. [Google Scholar] [CrossRef]
  56. Cahigas, M.M.; Prasetyo, Y.T.; Persada, S.F.; Ong, A.K.; Nadlifatin, R. Understanding the perceived behavior of public utility bus passengers during the era of COVID-19 pandemic in the Philippines: Application of social exchange theory and theory of planned behavior. Res. Transp. Bus. Manag. 2022, 2022, 100840. [Google Scholar] [CrossRef]
  57. Barclay, J.; Few, R.; Armijos, M.T.; Phillips, J.C.; Pyle, D.M.; Hicks, A.; Brown, S.K.; Robertson, R.E. Livelihoods, wellbeing and the risk to life during volcanic eruptions. Front. Earth Sci. 2019, 7, 2296–6463. [Google Scholar] [CrossRef]
  58. Warsini, S.; Buettner, P.; Mills, J.; West, C.; Usher, K. The psychosocial impact of the environmental damage caused by the Mt Merapi eruption on survivors in Indonesia. EcoHealth 2014, 11, 491–501. [Google Scholar] [CrossRef]
  59. Hershkovich, O.; Gilad, D.; Zimlichman, E.; Kreiss, Y. Effective medical leadership in times of emergency: A perspective. Disaster Mil. Med. 2016, 2, 4. [Google Scholar] [CrossRef] [PubMed]
  60. Martinez-Villegas, M.M.; Solidum, R.U.; Saludadez, J.A.; Pidlaoan, A.C.; Lamela, R.C. Moving for safety: A qualitative analysis of affected communities’ evacuation response during the 2014 Mayon Volcano Eruption. J. Appl. Volcanol. 2021, 10, 6. [Google Scholar] [CrossRef]
  61. Gaillard, J.-C. Alternative paradigms of volcanic risk perception: The case of Mt. Pinatubo in the Philippines. J. Volcanol. Geotherm. Res. 2008, 172, 315–328. [Google Scholar] [CrossRef]
  62. Baxter, P.J.; Aspinall, W.P.; Neri, A.; Zuccaro, G.; Spence, R.J.S.; Cioni, R.; Woo, G. Emergency planning and mitigation at vesuvius: A new evidence-based approach. J. Volcanol. Geotherm. Res. 2008, 178, 454–473. [Google Scholar] [CrossRef]
  63. Perry, R.W.; Lindell, M.K. Volcanic risk perception and adjustment in a multi-hazard environment. SSRN Electron. J. 2008, 172, 170–178. [Google Scholar]
  64. Morganstein, J.C.; Ursano, R.J. Ecological disasters and mental health: Causes, consequences, and interventions. Front. Psychiatry 2020, 11, 1. [Google Scholar] [CrossRef] [PubMed]
  65. Abella, A.A.; Prasetyo, Y.T.; Young, M.N.; Nadlifatin, R.; Persada, S.F.; Perwira Redi, A.A.; Chuenyindee, T. The effect of positive reinforcement of behavioral-based safety on safety participation in Philippine coal-fired power plant workers: A partial least square structural equation modeling (PLS-SEM) approach. Int. J. Occup. Saf. Ergon. 2022, 1, 1–27. [Google Scholar] [CrossRef] [PubMed]
  66. Kusumastuti, R.D.; Arviansyah, A.; Nurmala, N.; Wibowo, S.S. Knowledge management and natural disaster preparedness: A systematic literature review and a case study of East Lombok, Indonesia. Int. J. Disaster Risk Reduct. 2021, 58, 102223. [Google Scholar] [CrossRef]
  67. Hanel, P.H.; Maio, G.R.; Soares, A.K.; Vione, K.C.; de Holanda Coelho, G.L.; Gouveia, V.V.; Patil, A.C.; Kamble, S.V.; Manstead, A.S. Cross-cultural differences and similarities in human value instantiation. Front. Psychol. 2018, 9, 849. [Google Scholar] [CrossRef] [PubMed] [Green Version]
Sustainability 14 11329 g001 550
Figure 1. Philippine Geographic Location of Active and Potentially Active Volcanoes.
Figure 1. Philippine Geographic Location of Active and Potentially Active Volcanoes.
Sustainability 14 11329 g001
Sustainability 14 11329 g002 550
Figure 2. Conceptual Framework.
Figure 2. Conceptual Framework.
Sustainability 14 11329 g002
Sustainability 14 11329 g003 550
Figure 3. Classification Model from Random Forest Classifier. Legends: X0—Perceived Severity (PS); X1—Media (M); X2—Perceived Vulnerability (PV); X3—Intention (IN).
Figure 3. Classification Model from Random Forest Classifier. Legends: X0—Perceived Severity (PS); X1—Media (M); X2—Perceived Vulnerability (PV); X3—Intention (IN).
Sustainability 14 11329 g003
Sustainability 14 11329 g004 550
Figure 4. RFC Scatter Plot.
Figure 4. RFC Scatter Plot.
Sustainability 14 11329 g004
Sustainability 14 11329 g005 550
Figure 5. ANN Average Testing Accuracy Scatter Plot.
Figure 5. ANN Average Testing Accuracy Scatter Plot.
Sustainability 14 11329 g005
Sustainability 14 11329 g006 550
Figure 6. Training and Validation Loss Rate.
Figure 6. Training and Validation Loss Rate.
Sustainability 14 11329 g006
Sustainability 14 11329 g007 550
Figure 7. Optimum ANN Model.
Figure 7. Optimum ANN Model.
Sustainability 14 11329 g007
Sustainability 14 11329 g008 550
Figure 8. Taylor Diagram.
Figure 8. Taylor Diagram.
Sustainability 14 11329 g008
Sustainability 14 11329 g009 550
Figure 9. Violin Plot.
Figure 9. Violin Plot.
Sustainability 14 11329 g009
Sustainability 14 11329 g010 550
Figure 10. Box Plot.
Figure 10. Box Plot.
Sustainability 14 11329 g010
Table
Table 1. Indicators statistical analysis.
Table 1. Indicators statistical analysis.
VariableItemMeanStDCronbach’s Alpha
MediaMP14.3030.8330.803
MP24.3830.799
MP34.6630.614
MP44.3810.761
Hazard KnowledgeHK13.5121.1530.737
HK24.8090.485
HK34.7780.527
HK44.6430.698
HK54.6820.650
Perceived Risk ProximityGP12.3801.2600.773
GP22.7531.290
GP33.5991.142
Perceived SeverityPS14.0520.8400.829
PS24.4430.725
PS33.4900.993
PS43.4521.256
PS53.2991.256
Perceived VulnerabilityPV13.0541.2000.851
PV22.9251.214
PV32.5101.262
PV44.0841.100
Perceived Behavioral ControlPBC13.5121.0050.758
PBC23.5640.944
PBC33.6751.001
PBC43.5610.974
Social NormSN13.2161.0100.837
SN23.0061.104
SN32.8521.091
SN43.1531.066
SN53.5441.030
Attitude Toward the BehaviorATB13.3921.1070.796
ATB24.1380.965
ATB34.0920.993
ATB44.2830.820
ATB53.2761.047
IntentionIF14.4960.7020.844
IF24.1930.925
IF34.1330.933
IF44.1580.961
Perceived PreparednessPP13.6740.9750.822
PP24.1260.804
PP34.3310.746
PP44.4060.783
PP54.5540.687
Table
Table 2. Decision Tree Mean Accuracy (Depth = 5).
Table 2. Decision Tree Mean Accuracy (Depth = 5).
Category60:4070:3080:2090:10
Random
Gini88.0088.6090.0088.60
Std. Dev0.0003.8480.0000.548
Entropy90.2091.0086.6081.60
Std. Dev1.0960.0002.5101.140
Best
Gini92.0090.2093.0092.80
Std. Dev0.0000.8370.0002.280
Entropy89.6088.2092.0089.80
Std. Dev0.8943.1140.0001.483
Table
Table 3. Summary of ANN.
Table 3. Summary of ANN.
LatentAverage TrainingStandard DeviationAverage TestingStandard Deviation
Media96.113.13797.890.837
Perceived Severity91.303.22497.331.782
Perceived Vulnerability84.302.51597.171.286
Intention81.930.37796.951.839
Perceived Risk Proximity88.931.40095.652.133
Subjective Norm86.790.97995.342.173
Hazard Knowledge86.493.13087.106.200
Perceived Behavioral Control67.866.01583.445.752
Attitude66.042.34279.924.580
Table
Table 4. Score of Importance.
Table 4. Score of Importance.
LatentImportanceScore (%)
M0.202100
PS0.19998.6
PV0.19998.5
IN0.19697.2
PR0.19696.8
SN0.19094.3
HK0.18189.8
PBC0.16883.2
AT0.16179.9
Publisher’s Note: MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Share and Cite

MDPI and ACS Style

German, J.D.; Redi, A.A.N.P.; Ong, A.K.S.; Prasetyo, Y.T.; Sumera, V.L.M. Predicting Factors Affecting Preparedness of Volcanic Eruption for a Sustainable Community: A Case Study in the Philippines. Sustainability 2022, 14, 11329. https://doi.org/10.3390/su141811329

AMA Style

German JD, Redi AANP, Ong AKS, Prasetyo YT, Sumera VLM. Predicting Factors Affecting Preparedness of Volcanic Eruption for a Sustainable Community: A Case Study in the Philippines. Sustainability. 2022; 14(18):11329. https://doi.org/10.3390/su141811329

Chicago/Turabian Style

German, Josephine D., Anak Agung Ngurah Perwira Redi, Ardvin Kester S. Ong, Yogi Tri Prasetyo, and Vince Louis M. Sumera. 2022. "Predicting Factors Affecting Preparedness of Volcanic Eruption for a Sustainable Community: A Case Study in the Philippines" Sustainability 14, no. 18: 11329. https://doi.org/10.3390/su141811329

Note that from the first issue of 2016, this journal uses article numbers instead of page numbers. See further details here.

Article Metrics

Back to TopTop