International Journal of Innovative Technology and Exploring Engineering (IJITEE)
ISSN: 2278-3075, Volume-9 Issue-2, December 2019
4274
Published By:
Blue Eyes Intelligence Engineering
& Sciences Publication
Retrieval Number: B7725129219/2019©BEIESP
DOI: 10.35940/ijitee.B7725.129219
Abstract: Landslide is one of the major natural hazards which
is experienced all over the world and causes huge losses to land
and property. Most of the landslides are generally caused by
multiple factors which act together to destabilize the slope. But
among them, the most common trigger for the landslides have
been excessive rainfall and no proper planning have been leading
to disastrous outcomes. So, in this research, mostly focus on the
landslides which are induced due to rainfall to find a solution to
the problem. It is present an overview on the challenges being
faced in the prediction of Rainfall induced landslides. Also the
objective is to find relevant approaches and techniques and judge
the best possible method and algorithms which gives the most
accurate results.
Keywords : Support Vectors, Probability, Prediction model,
Regression, Risk, Accuracy
I. INTRODUCTION
Very minimal importance is being given to the topic and
very less planning is being done in advance regarding it. So,
there is a need to find a reliable system through which the
occurrence of such hazards could be predicted and the losses
imposed because of it could be minimized.
There are already some models which have been proposed
but each has some liabilities as well as advantages. There has
not been a model which can prove to be fully reliable in such
adverse circumstances. The main goal through this research is
to analyze the various methods already in use and find out the
important functionalities within each system [10].
And today the most widely used method for the prediction
related topic is Machine Learning. Machine learning is a way
of identifying patterns in data and using them to automatically
make predictions or decisions in the future [1].
In this research, to analyzing the various methods and
Machine Learning algorithms which have already being
implemented and bring out the specific details regarding each
model for the Landslide Prediction and also attempting to
create a system which can prove to be reliable as well as
efficient under such circumstances.
Revised Manuscript Received on December 05, 2019.
* Correspondence Author
K.Uma*,School of Information Technology and Engineering, VIT
University, Vellore, India, drumakphd@gmail.com
C.Ramesh Kumar, School of Computing Science and Engineering,
Galgotias University, Uttar Pradesh, India.
mail: c.ramesh@galgotiasuniversity.edu.in
T.R.Saravanan, Department of CSE, Jeppiaar SRR Engineering college,
Chennai-603103, saravanan5_t_r@yahoo.co.in
M.Basha Khaja, Wipro Technology, Software engineer, Ireland, United
Kingdom, ghaja.bms@gmail.com
II. LITERATURE SURVEY
A significant number of papers have been reviewed
covering the landslide studies from different regions of the
world. There are some distinct, different and new
methodologies for the problem and each specific approach
has its own limitations and advantages. Each journal has been
divided into separate tables with column headings containing
the methods/Algorithms Used, Factors considered in the
Dataset, the final Results and findings of their work, and the
final Remarks [13][14].
A. Comparison Analysis
A wide range of models and methods have been used in the
various papers according to the dataset which was planned to
be used. Some of them had comparisons between the various
possible algorithms for finding out the best possible option for
their model. But the results were found very contrasting to
each other. According to some Support Vector Machines
(SVM) was the model giving best probabilities [1].The tree
based models such as Logistic Model Tree[2] and Random
Forest[4] were also proved to be better in some papers and
for some it was Regression and Naïve Bayes giving the best
results . Besides these some hybrid algorithms [3][5] have
also been used which according to their results were proven to
be better than the conventional methods which are mostly
used .
Besides these methods, the conventional methods like
Artificial Neural Networks (ANN) [8] and Multilayer
Perceptron’s [6] are also being used. But these methods were
majorly used when one of the factors among the dataset was
the Plane Curvature, and were mostly used for the prediction
of Slope Instabilities considering the Rate of Displacement of
the land as inputs.
Some other unique models such as Rotation Forest Fuzzy
Rules Based Classifier Ensemble (RFCE) [7], Wavelet
Transform (WT) and the Artificial Bees colony (ABC) were
also used but the details about their implementation were not
shared.
Every research had their unique set of factors considered
for the dataset. Among them, Rainfall and Slope were found
to be the most common factors. The factors were chosen
taking into consideration as to which location the data has
been collected. The weather and climate conditions play a
vital role in choosing the factors for a particular place or
location. This has been the reason for getting so much
disparity among the factors being considered in the various
research papers. Some considered Rainfall, Pressure, Wind
Speed in the dataset [1],
whereas in some Elevation,
Distance from River and Water
Prediction of Rainfall Induced Landslides using
Machine Learning Algorithms
K.Uma, C.RameshKumar, T.R.Saravanan, M.Basha Khaja