Professional Documents
Culture Documents
net/publication/333498072
CITATIONS READS
22 186
2 authors:
All content following this page was uploaded by Biswajeet Pradhan on 10 November 2021.
ABSTRACT Rockfall source identification is the most challenging task in rockfall hazard and risk assess-
ment. This difficulty rises in the areas where there is a presence of other types of the landslide, such as shallow
landslide and debris flow. The aim of this paper is to develop and test a hybrid model that can accurately
identify the source areas. High-resolution light detection and ranging data (LiDAR) was employed to derive
the digital terrain model (DTM), from which several conditioning factors were extracted. These conditioning
factors were optimized utilizing the ant colony optimization (ACO). Different machine learning algorithms,
namely, logistic regression (LR), random tree (RT), random forest (RF), support vector machine (SVM), and
artificial neural network (ANN), in addition to their ensemble models (stacking, bagging, and voting), were
examined. This is based on the selected best subset of conditioning factors and inventory dataset. Stacking
LR-RT (the best fit model) was then utilized to produce the probabilities of different landslide types. On the
other hand, the Gaussian mixture model (GMM) was optimized and applied for automatically identifying the
slope threshold of the occurrence of the different landslide types. In order to reduce the model sensitivity to
the alteration in various conditioning factors and to improve the model computations performance, land use
probability area was formed. The rockfall sources were identified by integrating the probability maps and
the reclassified slope raster based on the GMM results. The accuracy assessment reveals that the developed
hybrid model can identify the probable rockfall regions with an accuracy of 0.95 based on the validation
dataset and 0.94 on based the training dataset. The slope thresholds calculated by GMM were found to
be > 58◦ , 22◦ –58◦ , and 9◦ –22◦ for rockfall, shallow landslide, and debris flow, respectively. This indicates
that the model can be generalized and replicated in different regions, and the proposed method can be applied
in various landslides studies.
INDEX TERMS Remote sensing, machine learning, rockfall, hybrid model, GIS, LiDAR.
2169-3536 2019 IEEE. Translations and content mining are permitted for academic research only.
74570 Personal use is also permitted, but republication/redistribution requires IEEE permission. VOLUME 7, 2019
See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
A. M. Fanos, B. Pradhan: Spatial Ensemble Model for Rockfall Source Identification
in the assessment of rockfall hazard [7]. Rockfall source Despite the numerous attempts that have been conducted
regions, that correspond to the detachment regions of rocks, for rockfall source identification using photogrammetry or
are normally taken from evidences such as scree deposits at laser scanning data in local scale [21]–[23], and the few
foot-slope and talus slope [8], historical and field inventory studies in regional scale [4], There is a main problem which
of fallen rocks. Therefore, a DTM-based geomorphomet- is still not considered. This problem is when the study area
ric methods become more appropriate for identification of contains other types of landslides that have nearly comparable
probable rockfall sources [6]. The potential rockfall sources geomorphometric characteristics such as shallow landslide,
are identified using the slope angle distribution obtained rockfall, and debris flow. Nevertheless, Fanos [24] proposed
from high-resolution DTM crossed with other information a model for rockfall source identification using a single
obtained from topographic and geological maps as GIS lay- machine learning algorithm. However, their study did not
ers. The distribution of slope angle can be decomposed in take into account the optimization of the algorithm hyperpa-
many Gaussian distributions that can be considered as mor- rameters that can significantly influence the derived results.
phological units characteristics [9]. Moreover, they used a limited conditioning factors without
The steep slope angle is one of the major elements optimization or examining the multicollinearity among these
necessary for rockfall initiation [10]. The processes of factors. In addition, the GMM was run without optimization
gravity-driven surfaces in mountainous areas are exceedingly and based on the inventory dataset not on the geomorpholog-
associated with the topography steepness and the relief mor- ical units of the slope. Therefore, this paper aims to develop
phology, and thus reflects these instabilities [11]. The stabil- and test an ensemble model using high-resolution LiDAR
ity slope angles rely on the type of rock and the mechanical data to solve such problem. This model is based on the
and geometrical properties of the discontinuities set [12]. combination of machine learning algorithms and Gaussian
Consequently, it can be considered that the distribution of Mixture Model (GMM). The former is to produce the prob-
slope angle reflects the relief type, such as plains, alluvial, ability maps of different landslide types (rockfall, shallow
glacial, etc., and the rocks mechanical properties. Therefore, landslide, and debris flow), and the latter is to determine the
various rock types and morphologies result in a range of slope thresholds of different landslide types. Kinta Valley,
slope angle values that are characteristic for a particular Malaysia was selected to perform the proposed ensemble
morphotectonic setting and expressed in the distribution of model.
slope angle [13].
In the past decades landslide probability, hazard, and
risk have been widely analyzed and explored. Neverthe- II. STUDY AREA
less, the selection of optimized conditioning factors in such Rockfall incidents are frequently occurred in Malaysia due
application remains a difficult mission [4]. The mapping to the high and steep terrain in addition to high density of
of landslide probability employs topological, hydrological, rainfall. Kinta Valley which is situated in the eastern part
geological, and environmental factors. Some researchers of Malaysia (Fig. 1), is one of the most hardly hit dis-
presume that the precision and accuracy of the produced prob- tricts in Malaysia which contributes to the economic and
ability maps increase by increasing the number of condition- tourism sectors. Approximately, the Kinta district has an area
ing factors. In contrast, other researchers have proved that a of 1,950 km2 . Geographically, the area is located approxi-
small number of conditioning factors are adequate to generate mately between the northeast corner (101◦ 80 33, 4◦ 390 0400 )
landslide probability map with a reasonable accuracy [14]. and the southwest corner (101◦ 30 3200 , 4◦ 310 3100 ).
Depending on the aim, there are numerous categories of The geological setting of Kinta Valley and adjacent regions
data mining tasks. Classification is one of those categories. are limestone bedrock, granitic hills, and mine. Consequently,
The main aim of classification is to gain knowledge of hidden Kinta Valley and its surrounding areas have encountered
patterns to make prediction about the class of some unknown several engineering geologic problems including rockfall,
data [15], [16]. For data classification, most of the standard landslide, and debris flow. The limestone bedrocks rise over
machine learning algorithms can be utilized effectively for the alluvial plain in every region forming limestone hills with
classification precision if class labels are equally scattered. very steep slopes (sub- vertical to vertical). The study area
However, such algorithms show less or poor learning per- consists of diverse lithology with a high coverage of igneous
formance in case of classifying the imbalanced dataset that rocks. Such features are existing in the high-altitude areas
have variation in the class labels [17]. On the other hand, on the east and west sides of Kinta Valley. The sedimen-
some new methods were proposed for classification issues tary (limestone) and metamorphic rocks (marble) and are
such as robust manifold matrix factorization [18] and deep widely present in Kinta Valley [25].
learning [19]. In order to get the accuracy of classification Kinta Valley receives high precipitation along the year
algorithms, one or more algorithms can be combined and with average annual rainfall of 321 mm. In addition, it expe-
can get the reasonable accuracy. The process of combin- riences tropical climate with temperature ranging between
ing the multiple algorithms is called ensembling [18]. The 23 ◦ C and 35 ◦ C. The humidity in the study area is rela-
most popular ensemble models are voting, bagging, and tively high ∼81.5% (Meteorological Service Department of
stacking [20]. Malaysia).
IV. METHODOLOGY
This section presents an overview of the proposed hybrid
model to identify the potential rockfall sources using
high-resolution LiDAR dataset. The model is based on an
ensemble stacking LR-RT and GMM methods. The details
of this model and its implementation and validation methods
are given in the following sub-sections. The ensemble models
were implemented through Python, whereas the GMM and
ACO were implemented using MATLAB (2017).
A. OVERALL METHODOLOGY
The developing steps of the proposed model are illustrated
in Figure 3, which include four major steps. In the first stage,
field and remote sensing dataset were obtained. ALS dataset,
landslides inventory, and a GIS dataset that contains topo-
graphical, geological, rainfall, and vegetation were prepared.
The acquired datasets were pre-processed in the second step.
In addition, the conditioning factors were optimized using
ACO. The ACO was run with different subsets of the con-
ditioning factors to find the best subset for each landslides
types. Three various machine learning algorithms (RF, SVM,
and LR) were employed to validate and compare the obtained
results. The third stage was the core-processing module in the
FIGURE 2. (Continued.) Landslide conditioning factors. proposed method, which consisted of developing the stacking
LR - RT and GMM models. The stacking LR - RT model
was developed to produce a probability of landslide, rockfall,
with an accuracy of (91%). Field surveys were conducted to and debris flow occurrences in the area using the inventory
verify the land use map. A total number of twelve classes dataset and the best subset of the conditioning factors. The
were specified as shown in Figure 2. The lithology of Kinta stacking LR - RT model was employed after performing an
Valley is predominantly limestone / marble. Other types
include sandstone and granite are also presence in the study
area.
Finally, areas with low vegetation density are more
exposed to erosion and increase instability than highly vege-
tated regions. The low vegetation coverage can simply result
in landslides formation. In the current research, vegetation
density was utilized as a conditioning factor for mapping
landslide and rockfalls probabilities.
B. INVENTORY DATASET
The landslide inventory was obtained from different sources
involving historical records, field surveys, and remote sens-
ing. SPOT fused images and aerial photo (0.1 m) were uti-
lized for the visual identification of landslides in the selected
area. Field surveys were performed to map the landslides
can be occurred beneath vegetation or in regions that can-
not be detected from the satellite images. Multiple in-situ
campaigns were performed utilizing a GNSS device. A total
number of 179 samples with their related characteristics were
prepared for assessment (Figure 1). The inventory dataset
include three types of landslide namely rockfall, shallow FIGURE 3. The workflow of the proposed integrated model for detecting
landslides, and debris flow. The inventory was split into two potential rockfall sources.
elaborate comparative study with other individual ensemble producing probability map is to evaluate the significance of
models such as LR, RT, RF, SVM, and ANN and their each factor. The conditioning factors building is a hard mis-
ensembles (bagging, voting, and stacking). The selection sion and no particular rules exists to define the number of the
of the base models hyperparameters is significantly affect conditioning factors that adequate for a particular probability
the performance a model. In the current study, grid search assessment. Moreover, no frameworks exist for the chosen
method was used to select these parameters. The best model of conditioning factors. Such factors are normally selected
was selected based on the standard accuracy metrics such as based on the experts’ opinions [32].
overall accuracy, Receiver Operating Characteristics (ROC) In this research two conditioning factors datasets were
curves, and Prediction Rate Curve (PRC). Then the best fit built within the environment of GIS. The first dataset
ensemble model (stacking LR – RT) was trained based on were obtained from the airborne LiDAR data encompass-
the inventory dataset with the best subset of the conditioning ing of eight landslide conditioning factors: altitude, aspect,
factors and thus the landslide probability was generated for slope, curvature, sediment transport index (STI), topographic
each type. On the other hand, The GMM is developed to roughness index (TRI), topographic wetness index (TWI),
determine the distributions of slope angle and deriving the and stream power index (SPI). The second dataset was the
geomorphological units in the selected area. The GMM was addition of other conditioning factors: environmental and
first optimized to find the best number of Gaussian compo- geological factors of, land use/cover (LULC), distance from
nents. Consequently, GMM was run using the slope dataset road and river, soil, and geology. Seventeen conditioning
of the selected study areas to determine the slope thresh- factors were used for landslide probability mapping.
olds of each landslides types automatically. The final step The analysis of multicollinearity is a significant step in
composed of mapping, validations, and model comparisons landslide probability. The presence of a near-linear relation
with other models. Consequently, the produced landslides among factors can produce a division-by-zero problem during
probabilities were integrated with the reclassified slope based regression calculations. Such problem can result in abortion
on the obtained thresholds from GMM and then land-use of the calculations and inexact relationship; division by a very
probability area to detect the potential landslide, debris flow, small quantity still deforms the result. Consequently, it is sig-
and rockfall source areas. nificant to analyze the landslide conditioning factors before
probability mapping. The collinear (dependent) factors can be
B. DTM GENERATION identified through multicollinearity analysis by examining a
Filtering process is fundamental step to derive an accu- correlation matrix built via computing R2 . Several quantita-
rate Digital Terrain Model (DTM). This because of the raw tive approaches are existing for multicollinearity detection,
LiDAR data contain ground and up-ground points. Several such as examination the eigenvalue in a correlation matrix,
filtering methods are existed for generating DTM, and this evaluation the variance infiation factor (VIF), and pairwise
study employs two various algorithms: morphology-based scatter plots. In the current research, the VIF values were cal-
filtering and multi-scale comparison. The former is suitable culated to detect the multicollinearity for each conditioning
for terrain with small features and steep terrains whereas the factor. Moreover, communality similar to R2 was computed
later appropriate for urban area [28]. Simple Morphological for each factor.
Filter (SMRF) is a morphology-based method proposed by After multicollinearity was performed, Ant Colony Opti-
Pingel et al. [29]. The multi-resolution hierarchical classifi- mization (ACO) with three different types of machine learn-
cation (MHC) is a multi-scale comparison method developed ing algorithms (RF, SVM, and LR), was tested and compared.
by Chen et al. [30]. Generally, the advantage of the multiscale Consequently, the best fit method was utilized to identify the
comparison filtering methods is deriving an accurate DTM in best subset of landslides conditioning factors. The area under
urban regions with different man-made natural objects [28]. curve (AUC) and overall accuracy were used to assess the
Both SMRF and MHC were employed in this research to filter obtained results.
the point clouds of different regions of the study area. In last decade ACO has attracted an increasing num-
Once the point clouds were filtered by removing the non- ber of researchers, and its applications have developed
ground points, both Kriging and IDW were tested based on remarkably [33]. ACO was efficiently employed in sev-
cross-validation method. IDW was found as an appropriate eral remote sensing applications, such as optimal attribute
interpolation method due to the high density of the point subset selection [14], image segmentation [34], selection of
cloud [31]. Consequently, IDW interpolation method was parameters [35], and object derivation [36]. The advantages
used to generate the DTM based on the remaining points with of ACO are, first, it makes a probabilistic decision in terms
a resolution of 0.5 m. of local heuristic information and artificial pheromone trails
and consequently permits the examination of a huge solu-
C. OPTIMIZATION OF THE CONDITIONING FACTORS tions number than greedy heuristic [36]. Second, it assists
The variances between the factors characteristics should be the evaporation of pheromone trail, that is a process which
assessed to generate landslide probability maps that utilizes result in decreasing of pheromone trail intensity along time.
different conditioning factors. The conditioning factors char- The evaporation of pheromone aids to prevent the fast algo-
acteristics differ from area to another, thus, the first step in rithm convergence toward a suboptimal area [37]. Third, in
the rule exploring context, an ACO algorithm can imple- TABLE 1. The selected landslide conditioning factors.
ment a robust, flexible search for an optimal terms com-
bination (logical conditions) including the predictor factors
values [38]. Moreover, ACO provides better and more reli-
able results than genetic algorithm (GA) and the common
sequence method (CSM) [39], [40]. The overall workflow of
ACO-based attribute selection is presented in Figure. 4.
A GMM is a weighted sum of M component Gaussian optimized, and then penalize it with the parameters number
densities as given by the equation, in the model (the complexity of the model). Nevertheless,
M
the AIC penalizes for complexity less severely than BIC.
X Thus, the BIC tends to select simpler model that may underfit,
p(X/λ) = wi g (x/µi , 6i ) (2)
whereas AIC tends to select more complex model that may
i=1
overfit. Therefore, it is better to look at AIC and BIC criteria
In which x is a D-dimensional continuous-valued data to decide the best model. The best fit model is with lower AIC
vector (i.e. features or measurements), wi , i = 1, . . . .M, are or BIC values.
the mixture weights, and g (X /µi , 6i ) , i = 1, . . . , M are the The GMM was trained on slope dataset extracted from
component Gaussian densities. Each component density is a three different areas within the area of interest. Iterative
D-variate Gaussian function of the form, search based on AIC and BIC values was performed to select
X 1 the hyperparameters of the model. The output of the GMM
g X /µi , = P 1/2 model was thresholds of slope angle (MUs), which allows
i
(2π)D/2 i
detecting the probable sources of various landslide types
1 X−1
automatically.
exp − (X − µi )0 (X − µi ) (3)
2 i
For debris flow the slope angles ranged from 7◦ to 25◦ and
the average slope angle was 15.13◦ .
In addition, since the sampling points (landslide, debris
flow, and rockfall) are subjected to strong correlations in
different conditioning factors, the multicollinearity of the
factors was analyzed using the Variance Inflated Factor (VIF)
method. Table 2 shows the VIF values calculated among the
conditioning factors in the landslide, rockfall, and debris flow
data samples. According to the previous works [26], a VIF
value of greater than four is considered high collinear. Thus,
the corresponding factors should be removed from further
analysis. The highest VIF value is 3.107 for slope factor in
the landslide data samples, 3.272 for STI factor in the rockfall
data samples, and 3.155 for altitude factor in the debris flow
FIGURE 5. Assessment of the factors subset selected by ACO based on
data samples. As a result, none of the factors was removed. overall accuracy for (a) landslide, (b) rockfall, and (c) debris flow.
TABLE 3. Selected conditioning factors based on ACO for each percent subset for shallow landslide.
TABLE 4. Selected conditioning factors based on ACO for each percent subset for rockfall.
by ACO. It also illustrates the classification performance for shallow landslide, rockfall, and debris flow. The ideal
according to the overall accuracy. The highest overall accu- number of factors among 17 available factors is 14, 12, and 15
racy was observed in experiment 8, 7, and 9 for shallow for shallow landslide, rockfall, and debris flow, respectively.
landslide, rockfall, and debris flow, respectively. In which 14, Careful analysis of Table 3 indicates that many factors, such
12, and 15 factors were selected by the algorithm for shallow as slope, distance to road, and distance to river are strongly
landslide, rockfall, and debris flow, respectively. Accord- affect the shallow landslide probability. In addition, RF out-
ingly, this subset was considered the best factors subset for performs LR and SVM, respectively. However, with the three
the dataset and the study area. evaluators the best subset was in experiment 8 with 80 %
The selected factors in each experiment for shallow land- of conditioning factors (14 factors). Table 4 shows the best
slide, rockfall, and debris flow are shown in Table 3, Table 4, subset of factors for rockfall probability. Slope, distance to
and Table 5, respectively, and the factors names are presented lineament and TRI, are found highly important for rockfall
in Table 1. probability mapping. The best subset was observed in experi-
The result indicates that utilizing a big number of condi- ment 7 with 70 % of conditioning factors (12 factors). The
tioning factors does not necessarily increase the accuracy of highest overall accuracy and AUC were recorded with RF
the calcification. The experiments illustrate that the accuracy evaluator which are 86 and 0.89, respectively. The best factors
of classification deteriorated after experiment 8, 7, and 9 subset for debris flow is shown in Table 5. Factors such
FIGURE 6. Probability maps based on stacking LR – RT model of (a) shallow landslide, (b) rockfall, and (c) debris flow.
TABLE 5. Selected conditioning factors based on ACO for each percent subset for debris flow.
TABLE 6. Accuracy assessment of the individual algorithms for different landslide types.
as distance to road, TRI, and vegetation density are found conditioning factors (15 factors). LR outperforms RF and
strongly influence the probability of debris flow. Neverthe- SVM with overall accuracy of 80. However, based on AUC,
less, the best subset is in experiment 9 with 90 % of the RF outperforms LR and SVM, respectively.
The identification of probable rockfall sources was per- PRC, and overall accuracy. RF outperforms ANN, LR, SVM,
formed by crossing the obtained threshold of slope angle and RT, respectively. The highest accuracy (ROC = 0.89,
(> 58◦ ) and the probability values of the sacking LR - RT. PRC = 0.86, and overall accuracy = 85.62), was achieved
Figure 8 demonstrates the result of potential rockfall sources by RF with rockfall data samples. However, all algorithms
in the study areas. The inventories of rockfalls are mainly achieved the highest accuracy with rockfall data samples in
situated in the regions those detected as probable source areas comparison with shallow landslide and debris flow. On the
proving that the proposed model is eligible to identify the other hand, LR, RT, and ANN performed better with debris
sources of both rockfalls and landslides in the selected areas. flow data samples than shallow landslide. Whereas, RF and
SVM performed better with shallow landslide data sam-
G. THE ASSESSMENT OF THE MODEL ACCURACY ples than debris flow. The comparison of bagging models
In order to validate the proposed ensemble model for pre- is shown in Table 7. RF outperformed the other machine
dicting the probable landslides the success (ROC) and pre- learning algorithms. Nevertheless, RF in bagging model was
diction (PRC) rate curves were employed [47]. The ROC slightly improved in comparison with individual RF this is
is generated based on the training landslide dataset while because RF itself is ensemble model. On the other hand,
PRC was produced utilizing the validation landslide dataset. the bagging models of the other algorithms showed bet-
Moreover, the area under curve (AUC) is used to identify the ter improvement than the individual implementation of the
accuracy of the probability maps [49]–[54]. same algorithms. Table 8 demonstrates the comparison of
Table 6 illustrates the comparison among the various voting models. Nine voting ensemble models were tested
machine learning algorithms individually based on ROC, based on the selected machine learning algorithms using
FIGURE 8. The probable rockfall source areas and the inventory dataset at (a) Gua Tambun, (b) Gunung Rapat, and (c) Gunung Lang.
rockfall, shallow landslide, and debris flow data samples. LR – RT model was employed to produce the probability
The accuracy of some algorithms was improved by adding maps of shallow landslide, rockfall, and debris flow. Table 10
another algorithm while some the accuracy of some algo- shows the comparative result of the assessment of the pro-
rithms was decreased by adding another algorithm. For posed ensemble models in recent literature.
instance, the accuracy of LR was improved by adding RT, Since the stacking LR – RT model performs well on both
RF, and ANN, whereas the accuracy of RF was decreased datasets (training and validation), the proposed ensemble
by adding SVM and ANN. Table 9 illustrates the compar- model is thus considered effective for landslides probability
ison of different stacking models. The stacking LR – RT mapping. In addition, as the model achieved a good accuracy
model outperformed the other tested models with all land- on validation landslides, it is expected to perform well in other
slides data samples (shallow landslide, rockfall, and debris regions that have similar or nearly similar condition as the
flow). The highest accuracy that achieved using stacking tested area.
LR – RT model with rockfall data samples is 0.94, 0.95,
and 92.89 of ROC, PRC, and overall accuracy. However, H. FIELD VERIFICATION
the model achieved good accuracy with shallow landslide In-situ survey was carried out to verify the produced map of
and debris flow of 0.86, 0.88, and 86.46, and 0.84, 0.81, rockfall sources areas. Several locations were selected ran-
and 83.95 of ROC, PRC, and overall accuracy, respectively. domly that distributed on the entire study area. Figure 9 shows
In comparison with some other tested models in the literature, the selected locations and the in-situ photos. All locations
the stacking LR - RT model achieved better accuracy than were found reasonably prone to rockfall incidents.
a hybrid Bagging based Support Vector Machines (BSVM) Discontinuities and fractures are clear in the selected
model [54], Decision tree (DT)-based CHi-squared Auto- locations especially along the cliff face. In addition, some
matic Interaction Detection (CHAID) model [56], a hybrid inventory fallen rocks were observed in various locations
approach of Reduced Error Pruning Trees (REPT) and Ran- close to the foot-slope. According to different interviews that
dom Subspace Ensemble (RSS) [57], and a hybrid fuzzy performed in the study area, these areas were encountered
weight of evidence model [58]. Consequently, the stacking rockfall incidents in different periods of time.
[14] M. N. Jebur, B. Pradhan, and M. S. Tehrany, ‘‘Optimization of landslide [36] L.-L. Li and J.-K. Wang, ‘‘SAR image ship detection based on ant colony
conditioning factors using very high-resolution airborne laser scanning optimization,’’ in Proc. 5th Int. Congr. Image Signal Process., Oct. 2012,
(LiDAR) data at catchment scale,’’ Remote Sens. Environ., vol. 152, pp. 1100–1103.
pp. 150–165, Sep. 2014. [37] J. Gottlieb, M. Puchta, and C. Solnon, ‘‘A study of greedy, local search, and
[15] F. Luo, B. Du, L. Zhang, L. Zhang, and D. Tao, ‘‘Feature learning ant colony optimization approaches for car sequencing problems,’’ in Proc.
using spatial-spectral Hypergraph discriminant analysis for hyperspectral Workshops Appl. Evol. Comput. Berlin, Germany: Springer, Apr. 2003,
image,’’ IEEE Trans. Cybern., vol. 49, no. 7, pp. 2406–2419, Jul. 2019. pp. 246–257.
[16] L. Zhang, L. Zhang, B. Du, J. You, and D. Tao, ‘‘Hyperspectral image [38] M. Dorigo and T. Stützle, ‘‘The ant colony optimization metaheuristic:
unsupervised classification by robust manifold matrix factorization,’’ Inf. Algorithms, applications, and advances,’’ in Handbook of Metaheuristics.
Sci., vol. 485, pp. 154–169, Jun. 2019. Boston, MA, USA: Springer, 2003, pp. 250–285.
[17] Y. Gu, Q. Wang, X. Jia, and J. A. Benediktsson, ‘‘A novel MKL model [39] R. S. Parpinelli, H. S. Lopes, and A. A. Freitas, ‘‘Data mining with an ant
of integrating LiDAR data and MSI for urban area classification,’’ IEEE colony optimization algorithm,’’ IEEE Trans. Evol. Comput., vol. 6, no. 4,
Trans. Geosci. Remote Sens., vol. 53, no. 10, pp. 5312–5326, Oct. 2015. pp. 321–332, Aug. 2002.
[18] S. P. Potharaju and M. Sreedevi, ‘‘Ensembled rule based classification [40] R. Putha, L. Quadrifoglio, and E. Zechman, ‘‘Comparing ant colony
algorithms for predicting imbalanced kidney disease data,’’ J. Eng. Sci. optimization and genetic algorithm approaches for solving traffic sig-
Technol. Rev., vol. 9, no. 5, pp. 201–207, 2016. nal coordination under oversaturation conditions,’’ Comput. Aided Civil
[19] X. X. Zhu, D. Tuia, L. Mou, G.-S. Xia, L. Zhang, F. Xu, and F. Fraundorfer, Infrastruct. Eng., vol. 27, no. 1, pp. 14–28, 2012.
‘‘Deep learning in remote sensing: A comprehensive review and list of [41] Z. Jiang, M. Zhou, J. Tong, H. Jiang, Y. Yang, A. Wang, and Z. You, ‘‘Com-
resources,’’ IEEE Geosci. Remote Sens. Mag., vol. 5, no. 4, pp. 8–36, paring an ant colony algorithm with a genetic algorithm for replugging tour
Dec. 2017. planning of seedling transplanter,’’ Comput. Electron. Agricult., vol. 113,
[20] M. Akour, I. Alsmadi, and I. Alazzam, ‘‘Software fault proneness pre- pp. 225–233, Apr. 2015.
diction: A comparative study between bagging, boosting, and stacking [42] D. N. Kumar and M. J. Reddy, ‘‘Ant colony optimization for multi-purpose
ensemble and base learner methods,’’ Int. J. Data Anal. Techn. Strategies, reservoir operation,’’ Water Resour. Manage., vol. 20, no. 6, pp. 879–898,
vol. 9, no. 1, pp. 1–16, 2017. 2006.
[21] A. M. Fanos, B. Pradhan, A. A. Aziz, M. N. Jebur, and H.-J. Park, ‘‘Assess- [43] A. M. Youssef, H. R. Pourghasemi, Z. S. Pourtaghi, and M. M. Al-Katheeri,
ment of multi-scenario rockfall hazard based on mechanical parameters ‘‘Landslide susceptibility mapping using random forest, boosted regression
using high-resolution airborne laser scanning data and GIS in a tropical tree, classification and regression tree, and general linear models and
area,’’ Environ. Earth Sci., vol. 75, no. 15, p. 1129, Aug. 2016. comparison of their performance at Wadi Tayyah Basin, Asir Region, Saudi
[22] B. Pradhan and A. M. Fanos, ‘‘Application of LiDAR in rockfall hazard Arabia,’’ Landslides, vol. 13, no. 5, pp. 839–856, 2016.
assessment in tropical region,’’ in Laser Scanning Applications in Land- [44] B. T. Pham, B. Pradhan, D. T. Bui, I. Prakash, and M. B. Dholakia,
slide Assessment. Cham, Switzerland: Springer, 2017, pp. 323–359. ‘‘A comparative study of different machine learning methods for landslide
susceptibility assessment: A case study of Uttarakhand area (India),’’
[23] R. Muzzillo, L. Losasso, and F. Sdao, ‘‘Rockfall source areas assessment
Environ. Model. Softw., vol. 84, pp. 240–250, Oct. 2016.
in an area of the Pollino national park (Southern Italy),’’ in Proc. Int. Conf.
Comput. Sci. Appl. Cham, Switzerland: Springer, Jul. 2018, pp. 366–379. [45] L. Xiao, Y. Zhang, and G. Peng, ‘‘Landslide susceptibility assessment
using integrated deep learning algorithm along the China-Nepal highway,’’
[24] A. M. Fanos, B. Pradhan, S. Mansor, Z. M. Yusoff, and A. F. B. Abdullah,
Sensors, vol. 18, no. 12, p. 4436, 2018.
‘‘A hybrid model using machine learning methods and GIS for potential
[46] Z. Ouyang, X. Sun, J. Chen, D. Yue, and T. Zhang, ‘‘Multi-view stacking
rockfall source identification from airborne laser scanning data,’’ Land-
ensemble for power consumption anomaly detection in the context of
slides, vol. 15, no. 9, pp. 1833–1850, 2018.
industrial Internet of Things,’’ IEEE Access, vol. 6, pp. 9623–9631, 2018.
[25] B. Pradhan, M. H. Abokharima, M. N. Jebur, and M. S. Tehrany, ‘‘Land
[47] Y. Chen, T. T. Georgiou, and A. Tannenbaum, ‘‘Optimal transport for
subsidence susceptibility mapping at Kinta Valley (Malaysia) using the
Gaussian mixture models,’’ IEEE Access, vol. 7, pp. 6269–6278, 2019.
evidential belief function model in GIS,’’ Natural Hazards, vol. 73, no. 2,
[48] C. Michoud, M.-H. Derron, P. Horton, M. Jaboyedoff, F.-J. Baillifard,
pp. 1019–1042, 2014.
A. Loye, P. Nicolet, A. Pedrazzini, and A. Queyrel, ‘‘Rockfall hazard and
[26] Y. Wu, B. Jiang, and N. Lu, ‘‘A descriptor system approach for estimation
risk assessments along roads at a regional scale: Example in Swiss Alps,’’
of incipient faults with application to high-speed railway traction devices,’’
Natural Hazards Earth Syst. Sci., vol. 12, no. 3, pp. 615–629, Mar. 2012.
IEEE Trans. Syst., Man, Cybern. Syst., to be published.
[49] A. Loye, M. Jaboyedoff, and A. Pedrazzini, ‘‘Identification of potential
[27] H. Hong, B. Pradhan, M. I. Sameen, W. Chen, and C. Xu, ‘‘Spatial pre- rockfall source areas at a regional scale using a DEM-based geomor-
diction of rotational landslide using geographically weighted regression, phometric analysis,’’ Natural Hazards Earth Syst. Sci., vol. 9, no. 5,
logistic regression, and support vector machine models in Xing Guo area pp. 1643–1653, 2009.
(China),’’ Geomatics, Natural Hazards Risk, vol. 8, no. 2, pp. 1997–2022, [50] N. Intarawichian and S. Dasananda, ‘‘Frequency ratio model based land-
2017. slide susceptibility mapping in lower Mae Chaem watershed, Northern
[28] Z. Chen, B. Gao, and B. Devereux, ‘‘State-of-the-art: DTM generation Thailand,’’ Environ. Earth Sci., vol. 64, no. 8, pp. 2271–2285, 2011.
using airborne LIDAR data,’’ Sensors, vol. 17, no. 1, p. 150, 2017. [51] A. Kornejady, H. R. Pourghasemi, and S. F. Afzali, ‘‘Presentation of RFFR
[29] T. J. Pingel, K. C. Clarke, and W. A. McBride, ‘‘An improved simple new ensemble model for landslide susceptibility assessment in Iran,’’ in
morphological filter for the terrain classification of airborne LIDAR data,’’ Landslides: Theory, Practice and Modelling. Cham, Switzerland: Springer,
ISPRS J. Photogramm. Remote Sens., vol. 77, pp. 21–30, Mar. 2013. 2019, pp. 123–143.
[30] C. Chen, Y. Li, W. Li, and H. Dai, ‘‘A multiresolution hierarchical clas- [52] W. Chen, H. Shahabi, A. Shirzadi, T. Li, C. Guo, H. Hong, W. Li, D. Pan,
sification algorithm for filtering airborne LiDAR data,’’ ISPRS J. Pho- J. Hui, M. Ma, M. Xi, and B. B. Ahmad, ‘‘A novel ensemble approach
togramm. Remote Sens., vol. 82, pp. 1–9, Aug. 2013. of bivariate statistical-based logistic model tree classifier for landslide
[31] J. Li and A. D. Heap, ‘‘Spatial interpolation methods applied in the environ- susceptibility assessment,’’ Geocarto Int., vol. 33, no. 12, pp. 1398–1420,
mental sciences: A review,’’ Environ. Model. Softw., vol. 53, pp. 173–189, 2018.
Mar. 2014. [53] B. T. Pham, A. Shirzadi, D. T. Bui, I. Prakash, and M. B. Dholakia,
[32] W. M. Abdulwahid and B. Pradhan, ‘‘Landslide vulnerability and risk ‘‘A hybrid machine learning ensemble approach based on a radial basis
assessment for multi-hazard scenarios using airborne laser scanning data function neural network and rotation forest for landslide susceptibility
(LiDAR),’’ Landslides, vol. 14, no. 3, pp. 1057–1076, 2017. modeling: A case study in the Himalayan area, India,’’ Int. J. Sediment
[33] J. Shang, X. Wang, X. Wu, Y. Sun, Q. Ding, J.-X. Liu, and H. Zhang, Res., vol. 33, no. 2, pp. 157–170, 2018.
‘‘A review of ant colony optimization based methods for detecting epistatic [54] J. Mathew, V. K. Jha, and G. S. Rawat, ‘‘Landslide susceptibility zonation
interactions,’’ IEEE Access, vol. 7, pp. 13497–13509, 2019. mapping and its validation in part of Garhwal Lesser Himalaya, India,
[34] L.-Y. Cao and L.-Z. Xia, ‘‘An ant colony optimization approach for SAR using binary logistic regression analysis and receiver operating character-
image segmentation,’’ in Proc. Int. Conf. Wavelet Anal. Pattern Recognit., istic curve method,’’ Landslides, vol. 6, no. 1, pp. 17–26, 2009.
Nov. 2007, pp. 296–300. [55] H. R. Pourghasemi, A. G. Jirandeh, B. Pradhan, C. Xu, and C. Gokceoglu,
[35] H. B. Alwan and K. R. Ku-Mahamud, ‘‘Optimizing support vector machine ‘‘Landslide susceptibility mapping using support vector machine and GIS
parameters using continuous ant colony optimization,’’ in Proc. 7th Int. at the Golestan Province, Iran,’’ J. Earth Syst. Sci., vol. 122, no. 2,
Conf. Comput. Converg. Technol. (ICCCT), Dec. 2012, pp. 164–169. pp. 349–369, 2013.
[56] B. T. Pham, D. T. Bui, and I, Prakash, ‘‘Bagging based support vector BISWAJEET PRADHAN received the Habilita-
machines for spatial prediction of landslides,’’ Environ. Earth Sci., vol. 77, tion degree in remote sensing from the Dres-
no. 4, p. 146, Feb. 2018. den University of Technology, Germany, in 2011.
[57] O. F. Althuwaynee, B. Pradhan, H.-J. Park, and J. H. Lee, ‘‘A novel He is currently the Director of the Centre for
ensemble decision tree-based CHi-squared automatic interaction detection Advanced Modelling and Geospatial Information
(CHAID) and multivariate logistic regression models in landslide suscep- Systems (CAMGIS), Faculty of Engineering and
tibility mapping,’’ Landslides, vol. 11, no. 6, pp. 1063–1078, 2014. IT. He is also the Distinguished Professor with the
[58] B. T. Pham and I. Prakash, ‘‘A novel hybrid model of bagging-based Naïve
University of Technology Sydney. He is also an
Bayes Trees for landslide susceptibility assessment,’’ Bull. Eng. Geol.
internationally established Scientist in the fields
Environ., vol. 78, no. 3, pp. 1911–1925, 2017.
[59] H. Hong, I. Ilia, P. Tsangaratos, W. Chen, and C. Xu, ‘‘A hybrid fuzzy of geospatial information systems (GIS), remote
weight of evidence method in landslide susceptibility analysis on the sensing and image processing, complex modeling/geo-computing, machine
Wuyuan area, China,’’ Geomorphology, vol. 290, pp. 1–16, Aug. 2017. learning and soft-computing applications, natural hazards, and environmen-
tal modeling. Since 2015, he has been serving as the Ambassador Scientist
for the Alexander Humboldt Foundation, Germany. Out of his more than
450 articles, more than 400 have been published in science citation index
(SCI/SCIE) technical journals. He has authored eight books and 13 book
chapters. He was a recipient of the Alexander von Humboldt Fellowship
ALI MUTAR FANOS was born in Baghdad, from Germany. He received 55 awards in recognition of his excellence in
Iraq, in 1987. He received the bachelor’s degree teaching, service, and research, since 2006. He was also a recipient of the
in surveying engineering from the University of Alexander von Humboldt Research Fellowship from Germany. From 2016 to
Baghdad, Iraq, in 2008, and the M.Sc. degree 2018, he was listed as the World’s Most Highly Cited Researcher by Clarivate
in GIS and geomatic engineering from Univer- Analytics Report as one of the world’s most influential mind. In 2018,
siti Putra Malaysia (UPM), in 2016, where he he was awarded as the World Class Professor by the Ministry of Research,
is currently pursuing the Ph.D. degree in geo- Technology and Higher Education, Indonesia. He is also an Associate Editor
graphic information system (GIS) with the Faculty and an Editorial Member of more than eight ISI journals. He has widely
of Engineering. Since 2008, he has been a Senior travelled abroad, visiting more than 52 countries to present his research
Engineer with the State Commission on Survey, findings.
Ministry of Water Resources, in different projects in Iraq.