An updated machine learning algorithm has been created by the Indian Institute of Technology, Mandi, which increases the precision of predicting natural hazards. The issue of data imbalance in estimating the likelihood of a landslide occurring in a particular location is addressed by the suggested approach.
Dericks Praise Shukla, an associate professor in the school of civil and environmental engineering at IIT Mandi, and Sharad Kumar Gupta, a former research student there who is now employed at Tel Aviv University, came up with the answer (Israel).
Finding the most vulnerable places is the first step in estimating and reducing the risk of landslides. Using the use of Landslide Susceptibility Mapping (LSM), which considers slope, elevation, geology, soil type, proximity to faults, rivers, and faults, as well as historical landslide data, it is possible to determine the likelihood that a landslide would occur in a given region. The reliable prediction of landslide occurrences is made possible by the application of machine learning techniques. Yet, because they are so uncommon, there aren’t enough training data sets available.
The novel machine learning algorithm created by Shukla’s team utilizes the EasyEnsemble and BalanceCascade undersampling strategies to address the problem of this data imbalance. Data on the landslides that happened in the Mandakini River Basin in Uttarakhand, India’s northwest Himalayas between 2004 and 2017 was gathered by the researchers to train and validate the algorithm.
“This new ML method demonstrates the potential for new technologies to generate substantial breakthroughs in the field and emphasizes the significance of data balancing in ML models. It also emphasizes how important it is to have a lot of data to effectively train ML models, especially in situations like geohazards and natural catastrophes where the stakes are high and people’s safety is at risk, according to Shukla.
Comparing the suggested solution to other methods like Support Vector Machines and Artificial Neural Networks, LSM’s accuracy is improved. The team claims that this approach can be used to predict other natural hazards like floods, avalanches, and extreme weather conditions.
