Spatial Prediction of Rainfall-Induced Landslides Using Aggregating One-Dependence Estimators Classifier

In this study, the spatial prediction of rainfall-induced landslides at the Pauri Gahwal area, Uttarakhand, India has been done using Aggregating One-Dependence Estimators (AODE) classifier which has not been applied earlier for landslide problems. Historical landslide locations have been collated with a set of influencing factors for landslide spatial analysis. The performance of the AODE model has been assessed using statistical analyzing methods and receiver operating characteristic curve technique. The predictive capability of the AODE model has also been compared with other popular landslide models namely Support Vector Machines (SVM), Radial Basis Function Neural Network (ANN-RBF), Logistic Regression (LR), and Naïve Bayes (NB). The result of analysis illustrates that the AODE model has highest predictability, followed by the SVM model, the ANN-RBF model, the LR model, and the NB model, respectively. Thus AODE is a promising method for the development of better landslide susceptibility map for proper landslide hazard management.     

Personalized travel route recommendation using collaborative filtering based on GPS trajectories

Travelling is a critical component of daily life. With new technology, personalized travel route recommendations are possible and have become a new research area. A personalized travel route recommendation refers to plan an optimal travel route between two geographical locations, based on the road networks and users’ travel preferences. In this paper, we define users’ travel behaviours from their historical Global Positioning System (GPS) trajectories and propose two personalized travel route recommendation methods – collaborative travel route recommendation (CTRR) and an extended version of CTRR (CTRR+). Both methods consider users’ personal travel preferences based on their historical GPS trajectories. In this paper, we first estimate users’ travel behaviour frequencies by using collaborative filtering technique. A route with the maximum probability of a user’s travel behaviour is then generated based on the naïve Bayes model. The CTRR+ method improves the performances of CTRR by taking into account cold start users and integrating distance with the user travel behaviour probability. This paper also conducts some case studies based on a real GPS trajectory data set from Beijing, China. The experimental results show that the proposed CTRR and CTRR+ methods achieve better results for travel route recommendations compared with the shortest distance path method.     

Landslide susceptibility assessment using three bivariate models considering the new topo-hydrological factor: HAND

Abstract

The main objective of this study is to assess the relative contribution of the state-of-the-art topo-hydrological factor, known as height above the nearest drainage (HAND), to landslide susceptibility modellling using three novel statistical models: weights-of-evidence (WofE), index of entropy and certainty factor. In total, 12 landslide conditioning factors that affect the landslide incidence were used as input to the models in the Ziarat Watershed, Golestan Province, Iran. Landslide inventory was randomly divided into a ratio of 70:30 for training and validating the results of the models. The optimum combination of conditioning factors was identified using the principal components analysis (PCA) method. The results demonstrated that HAND is the defining factor among hydrological and topographical factors in the study area. Additionally, the WofE model had the highest prediction capability (AUPRC = 74.31%). Therefore, HAND was found to be a promising factor for landslide susceptibility mapping.     

Evaluation and comparison of LogitBoost Ensemble,Fisher’s Linear Discriminant Analysis,logistic regression and support vector machines methods for landslide susceptibility mapping

The purpose of this study was to investigate and compare the capabilities of four machine learning methods namely LogitBoost Ensemble (LBE), Fisher’s Linear Discriminate Analysis (FLDA), Logistic Regression (LR) and Support Vector Machines (SVM) to select the best method for landslide susceptibility mapping. A part of landslide prone area of Tehri Garhwal district of Uttarakhand state, India, was selected as a case study. Validation of models was carried out using statistical analysis, the chi square test and the Receiver Operating Characteristic (ROC) curve. Result analysis shows that the LBE has the highest prediction ability (AUC = 0.972) for landslide susceptibility mapping, followed by the SVM (0.945), the LR (0.873) and the FLDA (0.870), respectively. Therefore, the LBE is the best and a promising method in comparison to other three models for landslide susceptibility mapping.     

A novel ensemble approach of bivariate statistical-based logistic model tree classifier for landslide susceptibility assessment

Abstract

This study addresses landslide susceptibility mapping (LSM) using a novel ensemble approach of using a bivariate statistical method (weights of evidence [WoE] and evidential belief function [EBF])-based logistic model tree (LMT) classifier. The performance and prediction capability of the ensemble models were assessed using the area under the ROC curve (AUROC), standard error, 95% confidence intervals and significance level P. Model performance analyses indicated that the AUROC values of the WoE–LMT ensemble model using the training and validation data-sets were 86.02 and 85.9%, respectively, whereas those of the EBF–LMT ensemble model were 88.2 and 87.8%, respectively. On the other hand, the AUC curves for the four landslide susceptibility maps indicated that the AUC values of the ensemble models of WoE–LMT (85.11 and 83.98%) and EBF–LMT (86.21 and 85.23%) could improve the performance and prediction accuracy of single WoE (84.23 and 82.46%) and EBF (85.39 and 81.33%) models for the training and validation data-sets.     

Effects of Training Samples and Classifiers on Classification of Landsat-8 Imagery

In this study, we used Landsat-8 imagery to test object- and pixel-based image classification approaches in an urban fringe area. For object-based classification, we applied four machine learning classifiers: decision tree (DT), naive Bayes (NB), random trees (RT), and support vector machine (SVM). For pixel-based classification, we utilized the maximum likelihood classifier (MLC). Specifically, we explored the influence of repeated sampling on classification results with different training sample sizes. We found that (1) except the overall accuracy of NB, those of the other four classifiers increased as the training sample size increased; (2) repeated sampling had a significant effect on classification accuracy, especially for the DT and NB classifiers; and (3) SVM achieved the best classification accuracy. In addition, the performance of the object-based classifiers was superior to that of the pixel-based classifier. The results of this study can provide guidance on the training sample size and classifier selection.     

A Comparative Study of Various Supervised Learning Approaches to Selective Omission in a Road Network

Selective omission is necessary for road network generalisation. This study investigates the use of supervised learning approaches for selective omission in a road network. To be specific, at first, the properties to measure the importance of a road in the network are viewed as input attributes, and the decision of such a road is retained or not at a specific scale is viewed as an output class; then, a number of samples with known input and output are used to train a classifier; finally, this classifier can be used to determine whether other roads to be retained or not. In this study, a total of nine supervised learning approaches, i.e., ID3, C4·5, CRT, Random Tree, support vector machine (SVM), naive Bayes (NB), K-nearest neighbour (KNN), multilayer perception (MP) and binary logistic regression (BLR), are applied to three road networks for selective omission. The performances of these approaches are evaluated by both quantitative assessment and visual inception. Results show that: (1) in most cases, these approaches are effective and their classification accuracy is between 70% and 90%; (2) most of these approaches have similar performances, and they do not have any statistically significant difference; (3) but sometimes, ID3 and BLR performs significantly better than NB and SVM; NB and KNN perform significantly worse than MP, SVM and BLR.     

基于随机森林算法的多维情境特征活动识别

利用智能手机传感器可感知时间、空间、时空和用户等多维情境的特征,可识别用户活动,但原框架模型中仅利用了单一分类器中的朴素贝叶斯算法,存在分类精度效果受限的问题。本文利用集成分类器中的随机森林算法对原有框架中的单一分类器进行了改进。在获取的3个数据集上的十倍交叉验证结果表明,加权平均F1量测值均有较大提高,表明利用随机森林算法在分类精度效果上有所提升;但由于集成算法结构相对复杂,其学习效率相对较低。此外,随机森林算法的分类混淆矩阵表明,导致识别误差的因素主要为活动的定义与室内定位精度。     

A novel hybrid approach of Bayesian Logistic Regression and its ensembles for landslide susceptibility assessment

Abstract

A novel artificial intelligence approach of Bayesian Logistic Regression (BLR) and its ensembles [Random Subspace (RS), Adaboost (AB), Multiboost (MB) and Bagging] was introduced for landslide susceptibility mapping in a part of Kamyaran city in Kurdistan Province, Iran. A spatial database was generated which includes a total of 60 landslide locations and a set of conditioning factors tested by the Information Gain Ratio technique. Performance of these models was evaluated using the area under the ROC curve (AUROC) and statistical index-based methods. Results showed that the hybrid ensemble models could significantly improve the performance of the base classifier of BLR (AUROC?=?0.930). However, RS model (AUROC?=?0.975) had the highest performance in comparison to other landslide ensemble models, followed by Bagging (AUROC?=?0.972), MB (AUROC?=?0.970) and AB (AUROC?=?0.957) models, respectively.     

Object-based image analysis of suburban landscapes using Landsat-8 imagery

Many experiments of object-based image analysis have been conducted in remote sensing classification. However, they commonly used high-resolution imagery and rarely focused on suburban area. In this research, with the Landsat-8 imagery, classification of a suburban area via the object-based approach is achieved using four classifiers, including decision tree (DT), support vector machine (SVM), random trees (RT), and naive Bayes (NB). We performed feature selection at different sizes of segmentation scale and evaluated the effects of segmentation and tuning parameters within each classifier on classification accuracy. The results showed that the influence of shape on overall accuracy was greater than that of compactness, and a relatively low value of shape should be set with increasing scale size. For DT, the optimal maximum depth usually varied from 5 to 8. For SVM, the optimal gamma was less than or equal to 10^?2, and its optimal C was greater than or equal to 10². For RT, the optimal active variables was less than or equal to 4, and the optimal maximum tree number was greater than or equal to 30. Furthermore, although there was no statistically significant difference between some classification results produced using different classifiers, SVM has a slightly better performance.     

A new supervised classifier exploiting spectral-spatial information in the Bayesian framework

Conventional machine learning methods are often unable to achieve high degrees of accuracy when only spectral data are involved in the classification process. The main reason of that inaccuracy can be brought back to the omission of the spatial information in the classification. The present paper suggests a way to combine effectively the spectral and the spatial information and improve the classification’s accuracy. In practice, a Bayesian two-stage methodology is proposed embodying two enhancements: i) a geostatistical non-parametric classification approach, the universal indicator kriging and ii) the smooth multivariate kernel method. The former provides an informative prior, while the latter overcomes the assumption (often not true) of independence of the spectral data. The case study reports an application to land-cover classification in a study area located in the Apulia region (Southern Italy). The methodology performance in terms of overall accuracy was compared with five state-of-the-art methods, i.e. naïve Bayes, Random Forest, artificial neural networks, support vector machines and decision trees. It is shown that the proposed methodology outperforms all the compared methods and that even a severe reduction of the training set does not affect seriously the average accuracy of the presented method.     

顾及样本优化选择的多核支持向量机滑坡灾害易发性分析评价

滑坡灾害易发性分析评价对地质灾害的防治与管理具有重要意义。针对滑坡灾害样本选择策略,单核支持向量机多特征映射不合理的问题,本文提出顾及样本优化选择的多核支持向量机(multiple kernel support vector machine,MKSVM)滑坡灾害易发性分析评价方法。为了保证样本平衡性并提高负样本的合理性,采用相对频率比(relative frequency,RF)综合评价各状态对于滑坡灾害易发性影响的重要程度,实现各评价因子状态的合理划分;利用确定性系数法(certainty factor,CF)计算各评价因子各状态分级影响滑坡灾害的敏感性,并在此基础上进行加权求和得到各栅格单元的滑坡灾害易发性指数,在滑坡灾害易发性指数极低和低易发区内随机选择与滑坡灾害点数目一致的非滑坡灾害点作为负样本数据。利用MKSVM对各特征空间最优核函数进行线性组合,解决了单一核函数映射不合理的问题,提高了模型的分类准确率和预测精度。以湖南省湘西土家族苗族自治州为研究区,从滑坡灾害易发性分区图、分区统计及评价模型精度3个方面对CF样本策略的MKSVM模型、CF样本策略的单核SVM模型、随机样本策略的MKSVM模型、随机样本策略的单核SVM模型进行了对比分析。结果表明,4种模型的受试者工作特征曲线(receiver operating characteristic,ROC)下的面积(area under curve,AUC)分别为0.859、0.809、0.798、0.766,验证了CF样本策略的合理性、有效性及MKSVM模型的可靠性。     

Spatial prediction of landslide susceptibility by combining evidential belief function,logistic regression and logistic model tree

Abstract

In this study, we introduced novel hybrid of evidence believe function (EBF) with logistic regression (EBF-LR) and logistic model tree (EBF-LMT) for landslide susceptibility modelling. Fourteen conditioning factors were selected, including slope aspect, elevation, slope angle, profile curvature, plan curvature, topographic wetness index (TWI), stream sediment transport index (STI), stream power index (SPI), distance to rivers, distance to faults, distance to roads, lithology, normalized difference vegetation index (NDVI), and land use. The importance of factors was assessed using correlation attribute evaluation method. Finally, the performance of three models was evaluated using the area under the curve (AUC). The validation process indicated that the EBF-LMT model acquired the highest AUC for the training (84.7%) and validation (76.5%) datasets, followed by EBF-LR and EBF models. Our result also confirmed that combination of a decision tree-logistic regression-based algorithm with a bivariate statistical model lead to enhance the prediction power of individual landslide models.     

Optimizing landslide susceptibility mapping in the Kongtong District,NW China: comparing the subdivision criteria of factors

Abstract

The standards applied to reclassify landslide-conditioning factors differ among studies and may change the accuracy of identifying landslide-prone areas. Therefore, we identified two standards per factor (elevation, aspect, slope, proximity to roads and proximity to streams) from the existing literature and set them as predisposing criteria in this paper. In addition to the five factors, lithology represented by types and a landslide inventory map produced from field surveys were also used in mapping. Thirty-two landslide susceptibility maps were generated based on weights-of-evidence and evaluated using the relative operative characteristic method. The results show that the subdivision criteria of factors change the accuracy, with the success rate varying from 84.34% to 87.51%. The map with the highest value captures more landslides in relatively higher susceptibility classes and is therefore considered the optimal one. Ultimately, a simplified mode of combining subdivision criteria is proposed to simplify comparison.     

GIS支持下滑坡灾害空间预测方法研究

滑坡预测在防灾减灾工作中具有重要意义,它包括空间、时间预测两个方面。基于统计模型进行区域评价与空间预测是滑坡灾害研究的重要方向,但是预测结果往往依赖样本数量和空间分布等。本文以马来西亚金马伦高原为研究区,选择高程、坡度、坡向、地表曲率、构造、土地覆盖、地貌类型、道路和排水系统作为评价因子,探讨运用地理信息系统(GIS)和遥感(RS)获取与管理滑坡灾害信息,以及热带雨林地区湿热环境下滑坡空间预测的方法。支持向量机(SVM)和逻辑(Logistic)回归模型分别应用于滑坡空间预测,结果显示平均预测精度分别为95.9%和86.2%,SVM法具有较高的描述精度,值得推荐;同时,基于SVM模型的滑坡空间预测受样本影响较小,预测结果相对比较稳定,这对于滑坡灾害区域评价与预测的快速实现具有实际意义。     

A GIS-based comparative study of Dempster-Shafer,logistic regression and artificial neural network models for landslide susceptibility mapping

The main aim of present study is to compare three GIS-based models, namely Dempster–Shafer (DS), logistic regression (LR) and artificial neural network (ANN) models for landslide susceptibility mapping in the Shangzhou District of Shangluo City, Shaanxi Province, China. At First, landslide locations were identified by aerial photographs and supported by field surveys, and a total of 145 landslide locations were mapped in the study area. Subsequently, the landslide inventory was randomly divided into two parts (70/30) using Hawths Tools in ArcGIS 10.0 for training and validation purposes, respectively. In the present study, 14 landslide conditioning factors such as altitude, slope angle, slope aspect, topographic wetness index, sediment transport index, stream power index, plan curvature, profile curvature, lithology, rainfall, distance to rivers, distance to roads, distance to faults and normalized different vegetation index were used to detect the most susceptible areas. In the next step, landslide susceptible areas were mapped using the DS, LR and ANN models based on landslide conditioning factors. Finally, the accuracies of the landslide susceptibility maps produced from the three models were verified using the area under the curve (AUC). The validation results showed that the landslide susceptibility map generated by the ANN model has the highest training accuracy (73.19%), followed by the LR model (71.37%), and the DS model (66.42%). Similarly, the AUC plot for prediction accuracy presents that ANN model has the highest accuracy (69.62%), followed by the LR model (68.94%), and the DS model (61.39%). According to the validation results of the AUC curves, the map produced by these models exhibits the satisfactory properties.     

VGI and crowdsourced data credibility analysis using spam email detection techniques

Volunteered geographic information (VGI) can be considered a subset of crowdsourced data (CSD) and its popularity has recently increased in a number of application areas. Disaster management is one of its key application areas in which the benefits of VGI and CSD are potentially very high. However, quality issues such as credibility, reliability and relevance are limiting many of the advantages of utilising CSD. Credibility issues arise as CSD come from a variety of heterogeneous sources including both professionals and untrained citizens. VGI and CSD are also highly unstructured and the quality and metadata are often undocumented. In the 2011 Australian floods, the general public and disaster management administrators used the Ushahidi Crowd-mapping platform to extensively communicate flood-related information including hazards, evacuations, emergency services, road closures and property damage. This study assessed the credibility of the Australian Broadcasting Corporation’s Ushahidi CrowdMap dataset using a Naïve Bayesian network approach based on models commonly used in spam email detection systems. The results of the study reveal that the spam email detection approach is potentially useful for CSD credibility detection with an accuracy of over 90% using a forced classification methodology.     

Landslide susceptibility mapping of a catchment area using frequency ratio,fuzzy logic and multivariate logistic regression approaches

Geospatial database creation for landslide susceptibility mapping is often an almost inhibitive activity. This has been the reason that for quite some time landslide susceptibility analysis was modelled on the basis of spatially related factors. This paper presents the use of frequency ratio, fuzzy logic and multivariate regression models for landslide susceptibility mapping on Cameron catchment area, Malaysia, using a Geographic Information System (GIS) and remote sensing data. Landslide locations were identified in the study area from the interpretation of aerial photographs, high resolution satellite images, inventory reports and field surveys. Topographical, geological data and satellite images were collected, processed, and constructed into a spatial database using GIS and image processing tools. There were nine factors considered for landslide susceptibility mapping and the frequency ratio coefficient for each factor was computed. The factors chosen that influence landslide occurrence were: topographic slope, topographic aspect, topographic curvature and distance from drainage, all from the topographic database; lithology and distance from lineament, taken from the geologic database; land cover from TM satellite image; the vegetation index value from Landsat satellite images; and precipitation distribution from meteorological data. Using these factors the fuzzy membership values were calculated. Then fuzzy operators were applied to the fuzzy membership values for landslide susceptibility mapping. Further, multivariate logistic regression model was applied for the landslide susceptibility. Finally, the results of the analyses were verified using the landslide location data and compared with the frequency ratio, fuzzy logic and multivariate logistic regression models. The validation results showed that the frequency ratio model (accuracy is 89%) is better in prediction than fuzzy logic (accuracy is 84%) and logistic regression (accuracy is 85%) models. Results show that, among the fuzzy operators, in the case with “gamma” operator (λ = 0.9) showed the best accuracy (84%) while the case with “or” operator showed the worst accuracy (69%).     

Comparison of landslide susceptibility mapping based on statistical index,certainty factors,weights of evidence and evidential belief function models

The main aim of this study was to produce landslide susceptibility maps using statistical index (SI), certainty factors (CF), weights of evidence (WoE) and evidential belief function (EBF) models for the Long County, China. Firstly, a landslide inventory map, including a total of 171 landslides, was compiled on the basis of earlier reports, interpretation of aerial photographs and supported by extensive field surveys. Thereafter, all landslides were randomly separated into two data sets: 70% landslides (120 points) were selected for establishing the model and the remaining landslides (51 points) were used for validation purposes. Eleven landslide conditioning factors, such as slope aspect, slope angle, plan curvature, profile curvature, altitude, distance to faults, distance to roads, distance to rivers, lithology, NDVI and land use, were considered for landslide susceptibility mapping in this study. Then, the SI, CF, WoE and EBF models were used to produce the landslide susceptibility maps for the study area. Finally, the four models were validated using area under the curve (AUC) method. According to the validation results, the EBF model (AUC = 78.93%) has a higher prediction accuracy than the SI model (AUC = 77.72%), the WoE model (AUC = 77.62%) and the CF model (AUC = 77.72%). Similarly, the validation results also indicate that the EBF model has the highest training accuracy of 80.25%, followed by SI (79.80%), WoE (79.71%) and CF (79.67%) models.     

A Multiple SVM System for Classification of Hyperspectral Remote Sensing Data

With recent technological advances in remote sensing sensors and systems, very high-dimensional hyperspectral data are available for a better discrimination among different complex land-cover classes. However, the large number of spectral bands, but limited availability of training samples creates the problem of Hughes phenomenon or ‘curse of dimensionality’ in hyperspectral data sets. Moreover, these high numbers of bands are usually highly correlated. Because of these complexities of hyperspectral data, traditional classification strategies have often limited performance in classification of hyperspectral imagery. Referring to the limitation of single classifier in these situations, Multiple Classifier Systems (MCS) may have better performance than single classifier. This paper presents a new method for classification of hyperspectral data based on a band clustering strategy through a multiple Support Vector Machine system. The proposed method uses the band grouping process based on a modified mutual information strategy to split data into few band groups. After the band grouping step, the proposed algorithm aims at benefiting from the capabilities of SVM as classification method. So, the proposed approach applies SVM on each band group that is produced in a previous step. Finally, Naive Bayes (NB) as a classifier fusion method combines decisions of SVM classifiers. Experimental results on two common hyperspectral data sets show that the proposed method improves the classification accuracy in comparison with the standard SVM on entire bands of data and feature selection methods.