In a study by Neenu and Lakshm in 2016, they created a prediction model for landslide in Cherrapunjee region of Meghalaya using Support Vector Machine(SVM), a technique meant to have strong capability to predict landslide by forecasting rainfall dataset on rainfall analysis. They concluded that SVM proved to be an efficient technique to forecast landslide by predicting the rainfall in advance.
Poonam and Neelam as cited in a Journal in 2018, enhanced the study of landslide prone area in Varunavat Parvat, India through supervised analysis. The area is located beside the Himalaya, situated at the right side of river Bhagirathi at a height of 1150m. They prepared data including slope instability and amount of rainfall to create prediction models using Artificial Neural Networks (ANN), Support Vector Machines (SVM), and logistic regression. They classified the outcomes of their predictions into classes being low, medium, or yes, no. Their comparison shows that SVM perform well in terms of accuracy for considered factors and in some cases logistic regression is performing better than SVM but SVM model best fits the hyperplane which divide the landslide prone groups.
In 2015, Poonam, Vibhuti, and Shivani created a GIS based model for monitoring and prediction of landslide susceptibility. They pointed out in their study that landslides tend to occur at any point in time and can cause huge damages to human life and resources, but the advancement in GIS based applications has eased out working on spatial or geographical data and will provide a powerful tool to model the landslide hazards for their spatial analysis and prediction. They proposed a GIS based landslide monitoring and forecast system using sensors. Using k-means clustering and ID3 decision trees they proposed a system to efficient and timely generation of landslide alerts.
In 2009, Milos and Branislav used SVM and k-Nearest Neighbor(k-NN) algorithms trained upon expert based model of landslide susceptibility using a multi-criteria analysis. They weighted the influences of different input parameters using Analytical Hierarchy Process (AHP). Their parameters included elevation, slope angle, aspect, distance from flows, vegetation cover, lithogy, and rainfall to represent the natural factors of the slope stability. The study, using machine learning classifiers included pattern recognition algorithms performed through training and testing mode. The SVM classifier outperformed the accuracy of the k-NN and turned out as quite a convenient classifier for landslide susceptibility as it turned out to be more consistent and precise. The most important result in their study was revealing that small training sets are sufficient to reach very high accuracy. They proposed to use a wider area but a multi-fold case with sparse inputs is yet to be confirmed.
In 2018, Jiubin, Yuanxue, and Ming performed a study on optimisation algorithm for decision trees and the prediction horizon displacement of landslide monitoring. In their study, they have pointed out the feature importance of the different attributes from GIS. They created a feature selection model which pointed out the important features: monitoring point, water level drop speed, monthly rainfall, max water level, daily rainfall, mean temperature, extreme minimum pressure, extreme minimum temperature, etc.
In 2017, Shuangxi, Qing et. al. created a model for predicting the landslide deformation with a knowledge-guided approach based on multi-mode monitor data using a Support Vector Regression (SVR). Using sensitivity coefficients to reflect the sensitive degree induced by multiple influencing factors, then a k-means clustering was implemented to discover the mechanism knowledge rules and finally deformation was predicted using SVR under the guiding of priori rules. They have concluded that their proposed knowledge-guided SVR approach is superior to the conventional SVR as verified in comparative experiments using displacement monitoring GIS datasets.