Abstract:In this study, an active probability backpropagation neural network model (PBNNM) was built by training a backpropagation neural network (BPNN) to predict the probability distribution of the probabilistic seismic hazard analysis (PSHA) monthly. The four-layered BPNN framework was determined using training data that were obtained from an earthquake catalogue for the time period of 1990-2015 (Taiwan Standard Time, TST). The studied region was divided into 500 small grids, each 0.2 • × 0.2 • in size, which is app… Show more
“…A four-layered BPNN with two hidden layers was employed based on the error backpropagation (EBP) algorithm (Omatu et al 2018;Lin and Chiou 2019), which is a technology of artificial intelligence (AI), shown in Figure 5. The Levenberg-Marquardt algorithm (LMA) was used as the EBP algorithm because it is a more commonly used and well-established method than other EBP algorithms for use in general nonlinear problems (Lera and Pinzolas, 2002;Mammadli, 2017).…”
Two backpropagation neural network (BPNN) models were constructed to predict two historic geomagnetic storms that occurred in September 1999 and October 2003. The Disturbance storm time (Dst) indices from January 1, 1999, to December 31, 2014 (Coordinated Universal Time, UTC), were used as the training and test datasets for cross-validation in order to verify and validate the reliability and robustness of the two BPNN models, and yielded reasonable, predicted results. A large correlation coefficient (R) and low root mean square error (RMSE) were obtained, verifying the reliability of the two BPNN models. The predicted Dst indices can be provided for giving inputs in advance (i.e., any future time). Therefore, this analyzed method can serve as an excellent real-time prediction (RTP). To test the ability and possibility of the long-term prediction (LTP) obtained using the two BPNN models, the Dst indices were examined, which corresponded to two significant historic large geomagnetic storms that occurred in August 1972 and March 1989. For the both BPNN models, after evaluating their learning procedure, the time-dependence of LTP, the dependence of the predicted errors on the time period length of training datasets, and the variance by learning process, we found that they were stable models for the RTP and LTP.
“…A four-layered BPNN with two hidden layers was employed based on the error backpropagation (EBP) algorithm (Omatu et al 2018;Lin and Chiou 2019), which is a technology of artificial intelligence (AI), shown in Figure 5. The Levenberg-Marquardt algorithm (LMA) was used as the EBP algorithm because it is a more commonly used and well-established method than other EBP algorithms for use in general nonlinear problems (Lera and Pinzolas, 2002;Mammadli, 2017).…”
Two backpropagation neural network (BPNN) models were constructed to predict two historic geomagnetic storms that occurred in September 1999 and October 2003. The Disturbance storm time (Dst) indices from January 1, 1999, to December 31, 2014 (Coordinated Universal Time, UTC), were used as the training and test datasets for cross-validation in order to verify and validate the reliability and robustness of the two BPNN models, and yielded reasonable, predicted results. A large correlation coefficient (R) and low root mean square error (RMSE) were obtained, verifying the reliability of the two BPNN models. The predicted Dst indices can be provided for giving inputs in advance (i.e., any future time). Therefore, this analyzed method can serve as an excellent real-time prediction (RTP). To test the ability and possibility of the long-term prediction (LTP) obtained using the two BPNN models, the Dst indices were examined, which corresponded to two significant historic large geomagnetic storms that occurred in August 1972 and March 1989. For the both BPNN models, after evaluating their learning procedure, the time-dependence of LTP, the dependence of the predicted errors on the time period length of training datasets, and the variance by learning process, we found that they were stable models for the RTP and LTP.
“…5), where J = K = 10. Some researchers have indicated that a neural network with two hidden layers and few neurons can replace a network with numerous neurons in one hidden layer (Lin et al, 2018;Lin and Chiou, 2019;Chu et al, 2020). The training data were used to train the BPNN.…”
Section: Validation By Two Back-propagation Neural Network (Bpnn) Modelsmentioning
confidence: 99%
“…(2) For the first BPNN model, all the first principal eigenvalues of the B2DPCA (Fig. 3a), with random noise added in the range of 0-0.5 (Chen et al, 2013) as target outputs, and their corresponding grids forming a 20 × 30 input matrix of positions 1) in the studies of Lin et al (2018) and Lin & Chiou (2019) as the training inputs to form the training data. The largest principal eigenvalues were considered to indicate the TIDs.…”
Section: Validation By Two Back-propagation Neural Network (Bpnn) Modelsmentioning
confidence: 99%
“…Two BPNN models were constructed to predict the magnitudes of all of the principal eigenvalues of the B2DPCA to determine whether the largest first and second principal eigenvalues were associated with the TIDs. Detailed explanations of BPNNs are available in the publications of Lin et al (2018) and Lin and Chiou (2019). The BPNN models were constructed as follows:…”
Section: Validation By Two Back-propagation Neural Network (Bpnn) Modelsmentioning
A weak tsunami was induced by the 2016 Mw = 7.8 Sumatra earthquake, which occurred at 12:49 on March 2, 2016 (UTC). The epicenter was at 5.060°S, 94.170°E at a depth of 10 km. At 15.02 on March 2 (UTC), the weak tsunami (amplitude: 0.11 m) arrived at the station located at 10.40°S, 105.67°E. The largest first principal eigenvalue derived using the bilateral projection-based two-dimensional principal component analysis (B2DPCA) indicated a spatial traveling ionospheric disturbance (TID), which was caused by internal gravity waves (IGWs), at 13:20 on March 2 (UTC). The largest second principal eigenvalue represented another TID expanding to the southwest. The two largest principal eigenvalues were associated with the TIDs, which were also determined using two back-propagation neural network (BPNN) models and two convolutional neural network (CNN) models, called the BPNN-B2DPCA and CNN-B2DPCA methods, respectively. These two methods yielded the same results as the B2DPCA. Therefore, the robustness and reliability of the B2DPCA were validated.
“…In addition, the development of artificial intelligence (AI) technology offers more possibilities for scientific assessments of geo-hazards risks. Numerous machine learning models, such as decision tree (DT) [23], support vector machine (SVM) [24], artificial neural network (ANN) [25], BP-artificial neural network (BP-ANN) [26], and Bayesian network (BN) models [27], have been applied for geo-hazards risk assessments. Among them, the SVM model, which is an efficient and reliable AI algorithm, has a very strong nonlinear processing ability and is one of the significant methods in risk assessment [28].…”
This study proposes an objective and accurate geo-hazards risk assessment method to address the challenge of increasingly severe hazards around the world. Previous studies mostly began from the perspectives of hazard and vulnerability, ignoring the role of survey data at disaster sites during risk assessments. The random forests (RF) model was applied in this study. Combined with detailed data from hazard sites, a geo-hazards risk assessment model was constructed, with the two dimensions of disaster hazard and vulnerability, was constructed. We analyzed the spatial pattern characteristics and the internal patterns of disaster risk and discussed the risk controlling factors and their contributions. The results showed the following. (1) The RF model, when combined with hazard, vulnerability conditions, and detailed data from disaster sites, can be used to zone and verify regional geo-hazards risks, providing a method for point-to-surface disaster risk mapping. (2) The RF-based geo-hazards risk assessment results were relatively consistent with the evaluation results from the support vector machine (SVM) model, but the accuracy and stability of the RF model were higher. (3) This method can be used to avoid the subjectivity in determining the weights and threshold values for indexes and can calculate the contribution of each index to geo-hazards risks. INDEX TERMS Geo-hazards, random forests model, risk assessment, Shifang county, support vector machine model.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.