support vector machine (svm)

In the current study, the risk of landslides in the Zamkan Watershed, located in Kermanshah Province, was evaluated. Two machine learning models, Support Vector Machine (SVM), and Logistic Regression, were used to prepare a landslide susceptibility map. Toward this, 13 informational layers including elevation, slope, aspect, Melton ruggedness number, terrain convexity, stream length, valley depth, topographic wetness index, precipitation, geological formations, distance from rivers, distance from roads, and vegetation cover were utilized as independent variables. Approximately 70% of the watershed's landslide pixels were used for model training, and 30% for model validation. Model validation was performed using ROC curves. The results indicated the higher performance and accuracy of the radial basis function (RBF) kernel of the SVM model for generating landslide hazard maps in the study area. The area under the curve (AUC) for the RBF kernel was approximately 0.951 for model training and 0.944 for model testing. The results suggest that slope with a coefficient of 0.28, precipitation with a coefficient of 0.27, lithology with a coefficient of 0.26, and elevation with a coefficient of 0.22 are the main controlling factors for landslides occurrence in the Zamkan Watershed. Both the SVM model and logistic regression confirmed the deterministic effects of selected factors on landslides. About 35% of the study area as classified as highly susceptible to landslides, primarily in the eastern half of the watershed. Factors such as high elevation, steep slopes, heavy precipitation, and the Kazhdomi Formation's composition were identified as key contributors to this susceptibility.

One of the key priorities of the Ministry of Agriculture Jihad is the mapping of croplands to estimate crop acreage and annual yield. In recent decades, remote sensing technology has proven to be highly effective in estimating the extent of crop cultivation through the use of timely images and synchronized data with diverse spatial, temporal, and spectral resolutions, leveraging advanced machine-learning algorithms. This study presented a framework for crop mapping in Marvdasht, Fars Province, by utilizing time series of Landsat-8 satellite images and advanced machine-learning algorithms. The employed algorithms included Decision Tree (DT), Random Forest (RF), Rotation Forest (RoF), Support Vector Machine (SVM), and Dynamic Time Warping (DTW) analysis. The results indicated that the dynamic time warping and random forest methods outperformed others, achieving significantly higher accuracy (with an overall accuracy improvement of 10-12%) in generating the agricultural land-use map of the study area. Furthermore, this research demonstrated the effectiveness of Bands 2-5 of Landsat-8 satellite in confidently identifying all crops in this region using the mentioned methods.

Occurrence of wildfires in forests is one of the important environmental hazards. Remote sensing is one of the useful sources for detecting and monitoring wildfires. The purpose of this paper is to evaluate "during fire" image and "before and after fire" images from Landsat8 satellite in identifying fire areas using Support Vector Machine (SVM) and Random Forest (RF) classifiers. Based on the analysis of the output from the images of the Sacramento area in the state of California, it was found that RF classification method with an overall accuracy of 99.83%, compared to the SVM method with an overall accuracy of 99.53%, has a better ability to distinguish fire from non-fire areas. It should be noted that in both methods, the overall accuracy was considerable and indicated their desirability to wildfire detection. Moreover, the classification results with a “single image” input during a fire were better than the “difference image” input.

Estimating the biomass values in forests stands through remote sensing is important. It has been reported that the major reasons of uncertainty are the lack of concurrency in satellite data and field information as well as the use of global allometric equations for estimating the weight of biomass of forest trees inside the country. Minimizing the above problems and the investigation of data performance in developing appropriate model for estimating the forest biomass in the Bankoll region of Karazan District of Sirvan County in Ilam province using Sentinel-1 satellite data in 27th of June, 2017 was the main goal of this study. Average size of the trees crown in 53 rectangular plots related to the coppice growth form with dimensions of 30×30 mwhich during 23 may 2017 to 10 June 2017 through applying DGPS by RTK method have been implemented on the ground were entered in the process of estimation the value of biomass. The average harvested field biomass was 10.63 Mg ha-1. After extraction of radar features, those features which had the greatest correlation with the values of biomass were selected using genetic algorithm by two models including K-Nearest Neighbor (K-NN) regression and Support-Vector Regression (SVR), then the most appropriate combination was identified and the biomass values were modelled. Models were validated using 26 test plots. Correlation of features obtained from radar data and the value of biomass indicated that features of VH، Mean VV، Mean VV GLCM (Correlation) and Mean VH GLCM (Dissimilarity) had the greatest sensitivity towards the value of biomass. Using regression models indicated that SVR model (Relative RMSE of 0.08) was more precise compared with K-NN regression (relative RMSE of 0.10). The best combination in the use of K-NN regression model with a relative RMSE of almost 0.99 Mg ha-1 (equal to 10%) and the coefficient of determination (R2) of 0.22 and the best combination when using SVR model was a relative RMSE of 0.87 Mg ha-1 (equal to 8%) and the R2 of 0.14. The results indicated that the final models, obtained from the optimal features extracted from radar data in the wavelength of C band and used parametric and non-parametric regressional methods in this research, were not abled to improve the saturated effect in data for estimation of biomass in the studied forests and it was not resulted in presenting an estimating model with an acceptable accuracy.

Landslides are one of the most important geological hazards worldwide (Chen et al., 2018). Despite advances in science and technology, these events continue to result in economic, human, and environmental losses worldwide (Alimohammadlou, Najafi, & Yalcin, 2013). Globally, landslides cause about 1200 deaths and 3.5 billion dollars of loss each year (Zhang, Han, Han, Li, Zhang, & Wang, 2019). About 66 million people live in landslide-prone areas (Chen et al, 2018). Landslide susceptibility (LS) mapping is essential in delineating landslide prone areas in mountainous regions. Landslide susceptibility is the propensity of soil or rock to produce various types of landslides (Chalkias Ferentinou, & Polykretis, 2014). From the beginning of the 1970s, the interest of both geoscientists and engineering professionals in LS zonation and the increasing emphasis on the use of Geographic Information Systems (GIS) technology led to the development of many methods such as weights-of-evidence model (Karami, 2012; Wang, Guo, Li, He, & Wu, 2019) logistic regression (Pham, Pradhan, Bui, Prakash, & Dholakia, 2016; Raja, Çiçek, Türkoğlu, Aydin, & Kawasaki, 2017), artificial neural networks (Chauhan, Sharma, Arora, Gupta, 2010؛ Tsangaratos & Benardos, 2014), neuro-fuzzy (Aghdam, Varzandeh, & Pradhan, 2016; Lee, Hong, & Jung, 2017; Chen et al., 2019). Support vector machines model (SVM) is currently a new pattern recognition method based on statistical learning theory, which shows unique advantages in solving small sample, nonlinear and high dimensional pattern recognition problem (Yao, Tham, & Dai, 2008).Iran has vast mountainous areas that make up more than half of the country due to geological characteristics, seismicity, rainfall and climate change and topographic conditions are among the countries that have experienced numerous landslides. In the meantime, Aharchai basin is located in northwest of Iran. This basin is prone to landslide due to topographic and geological conditions, slopes and other factors. Such conditions reveal the necessity of zoning sensitivity and assessing the potential for landslides in planning and implementing development plans. In this study, SVM model and GIS technology are applied to the evaluation of the susceptibility of landslides for the Ahar-Chai Basin. In addition, the purpose of this study is to evaluate the performance of support vector machine algorithm functions.

Support Vector Machine (SVM) is currently a new pattern recognition method based on statistical learning theory, which shows unique advantages in solving small sample, nonlinear and high dimensional pattern recognition problems (Yao et al, 2008). SVM was first proposed by Vapnik (1995). Landslide inventory mapping is an important step in landslide susceptibility assessment. In this study, historical records, satellite images, field surveys, and Google Earth® were used to analyze landslide locations. Landslide conditioning factors used in the current study are slope angle, slope aspect, altitude, valley depth, NDVI, rainfall, distance to rivers, lithology, distance to faults, land use, topographic wetness index (TWI),  stream power index (SPI), plan curvature, and profile curvature. Using this DEM, slope degree, slope aspect, altitude, plan curvature, profile curvature, were produced. In this study, a Landsat/ETM+ satellite image and sentinell were used for the year 2017. The plan curvature map was produced using a system for automated geoscientific analyses (SAGA) GIS. In this research, a support vector machine with four types of kernel classifiers such as linear, polynomial, radial basis function (RBF) and sigmoid were used in GIS for landslide susceptibility mapping in the study area. A total of 200 landslides were mapped using satellite image and subsequently were verified through field checking. Validation is an essential part of landslide susceptibility and landslide susceptibility maps are meaningless without validation. At present, the area under curve (AUC) method is used by most scholars in the prediction capability of a landslide susceptibility model. The AUC displays the success rate and prediction rate percentage of the model and is obtained for both the training data and the validation data. In general, the larger the AUC value, the better is the model (Intarawichian & Dasananda 2011; Lee and Dan, 2005).

In this study, using the 70% of the landslides as training dataset, the success rate curve was drawn. Using the remaining 30% of the landsides as testing dataset, the prediction rate curve was also drawn. Susceptibility maps were verified and compared using the area under the curve (AUC) method. The success rate curve demonstrated that the AUC for the radial basis, sigmoid, polynomial and linear functions was 0.988, 0.938, 0.652 and 0.506, respectively, and the prediction rate curve showed that the AUC was 0.958, 0.928, 0.642 and 0.543, respectively. Furthermore, results showed that RBF function had the highest accuracy in comparison with other methods. Generally, the three methods showed reasonable accuracy in landslide susceptibility mapping. Results of this study can serve as guidelines to managers and policy makers regarding the prevention and mitigation of landslide hazards. Finally, the landslide susceptibility maps were reclassified into five susceptibility classes: very high, high, moderate, low and very low.

The validation results showed that success rates for four types of function models varied from 98% to 50%. Similarly, results of prediction rates showed that RBF (98%) and sigmoid (93%) functions performed better than other types of functions (polynomial = 65%, and linear = 50%). The result of zonation show that 26.61 % of study area were located in high and very high susceptibility classes.

Landslide susceptibility mapping (LSM) is a proper method to predicting landslide hazard risk in order to reducing its consequences. We prepared the LSM mapping of Kan basin by used of index of Entropy model and SVM-S. The validation of the produced maps is evaluated by used of the area under the curve ROC. Study area The study area is located on the north west of Tehran Province, between 35°46′ and 35°58′ N latitude and 51°10′ to 51°23′ E longitude. The area of the basin is 204.385 km2 (Figure. 2 (a)). Material and methods In the present study we used of 8 parameters consisting the distance from river, distance from fault, distance from road, land use, rainfall, aspect, slope, elevation, and lithology. Distance from the road was computed from the road at the interval 200 m using ArcGIS software (Fig.2f). The distance from road and distance from fault was calculated in same way (Figure. 2 d, h). The land use map has reclassified to 5 class (Figure.2g). The lithology parameter has been obtained by the reclassification of the geological map of Tehran at the scale of 1:100000 (Figure.2h). The digital elevation model (DEM) was extracted from the 1:50000 scale topographic maps. The parameters of slope degree (whit 5 classes), aspect layer were produced by used of the digital elevation model. (Figure.2c & d). The topography layer was reclassified into 5 class (Fig.2a). the introduced layers were used in this study according to the type of the models to produce of LSM maps. Entropy index Entropy is a measurement of the instability, imbalance, and uncertainty of a system (yang et al, 2010). The equations used to calculate the information coefficient dj representing the weight value for the parameter as a whole, are given as follows: pij=xij/(∑_(i=1)^m▒xij ) (1) (2) Ej= -k+∑_(i=1)^m▒pij ln⁡(pij) K=〖c(lnm)〗^(-1) (3) ((4 dj = EJ – 1 After calculating the total weight (wj) using Equation 5, the landslide risk of the case study is evaluated: (5) Hi = ∑_(i=1)^m▒xij In equation 5, H is the coefficient of landslide risk, wj is the final weigh of all the factors and Xij is the weight of each factor (zongji et al, 2010). The final landslide susceptibility map was prepared by the summation of weighted products of the secondarily parametric maps. H = (S×0/54) + (Df × 0/74) + (E × 0/082) + (Dr × 0/51) + (Dri × 0/51) + (A × 0/066) + (lu × 0/16) + (Lt × 0/0064) Support Vector Machine SVM algorithm as one of the most popular methods for solving regression problems has had significant results in landslide sensitivity zoning. consider a set of linear separable training vectors Xi (i = 1, 2, . . ., n). The training vectors consist of two classes, which are denoted as Yi = ±1. The goal of SVM is to search an n-dimensional hyperplane differentiating the two classes by their maximum gap. Mathematically, it can be expressed as: 1⁄2=∥w^2∥ (6) Y_i=((W.X_i )+)≥1 (7) A Lagrangian formulation is introduced to solve the problem (equation 8). Thus, the goal is now to minimize the Lagrangian L with to W and b and maximize with respect to λi. For this reason, we used of following equation: L=1/2∥w^2∥-∑_(i=1)^n▒Y^i ((W.X_i )+b)-1) (8) four types of SVM is existed: linear, polynomial, radial basis function (RBF) and sigmoid. The mathematical representation of each kernel (linear, polynomial, radial basis function, and sigmoid) is listed below, respectively: K (X_j 〖.X〗_i )= X_j^i.X_j K (X_j 〖.X〗_i )=(γ∙ X_j^i+r) ._ γ>0 K (X_j 〖.X〗_i )=e^(-γ〖(X_i-X_j)〗^2 ) ._ γ>0 tanh (γ .X_i^T. X_j+r) γ, d, and r are user-controlled parameters, as their correct definition significantly increases the accuracy of the SVM solution. In the present research we used of Sigmoid function. To measure the validation of the models, we used of a relative ROC by comparing the existing landslide location with the two landslide susceptibility maps. The success rate curves were obtained by used of the 70% training dataset (29 landslide locations). ROC curve (AUC) represents the quality of the probabilistic model (it is ability to predict the occurrence or nonoccurrence of an event). Result and discussion The area of the low, moderate, and high classes based on the SVM model were found to be 109.485 km2, 38.7 km2, and 56.2 km2, respectively, whereas based on landslide susceptibility map by used of index of entropy, the 118.175 of the study area has low susceptibility risk, and the moderate, and high susceptibility zones have the 41.2 km2, 45.02 km2 of the study area, respectively (Fig. 3). Based on the entropy model, the 8 numbers of the landslides points located on the high-risk zone and the 8 numbers of the landslide points located on the moderate risk zone and low risk zone have 10 of the landslide points. Based on the LSM map produced by the SVM-S model, the 13 numbers of the landslide points located on the high risk zone and the 5 number of the landslide point located on the moderate-risk zone. The ROC plot assessment reveals that the AUC in the susceptibility map based on the index of entropy model was 0.86 and the AUC in the susceptibility map based on the Logistic Regression model was 0.91 (Fig. 5). Conclusion The high-risk zone on the LSM map produced by the SVM model is located in the north east and the west and south of the basin and based on the LSM map produced by the Entropy model is located in the north east and the south of the basin. The LSM map has produced in a regional scale, so further study need be carried out at the site-specific level to determine the exact extent site of the slope instability. Keywords: LSM, Index of Entropy model, Kan basin, Support Vector Machine, Sigmoid function, SVM-SIGMOID.