-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathpublications.html
44 lines (43 loc) · 4.09 KB
/
publications.html
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
<!DOCTYPE html>
<html>
<head><meta http-equiv="Content-Type" content="text/html; charset=windows-1252">
<title>Kapil Agrawal</title>
<link href="master.css" rel="stylesheet" type="text/css">
<script src="jquery-3.1.1.min.js.download" integrity="sha256-hVVnYaiADRTO2PzUGmuLJr8BLUSjGIZsDYGmIJLv2b8=" crossorigin="anonymous"></script>
<script type="text/javascript" src="navigation.js.download"></script>
</head>
<body style=""><link href="bootstrap.min.css" rel="stylesheet">
<link href="simple-sidebar.css" rel="stylesheet">
<script type="text/javascript" src="bootstrap.min.js.download"></script>
<div id="wrapper">
<div id="sidebar-wrapper">
<ul class="sidebar-nav">
<li class="sidebar-brand-k"><a href="index.html"><img class="img-circle" src="kitten.jpg"></a></li>
<li><a href="resume.pdf">CV</a></li>
<li><a href="publications.html">Publications</a></li>
<li><a href="projects.html">Projects</a></li>
<li><h4>Code</h4></li>
<li class="sublink"><a href="https://github.com/kapilagrawal95/smote_variants">SMOTE-IPF</a></li>
<li class="sublink"><a href="https://github.com/kapilagrawal95/spider">SPIDER-2</a></li>
<li><h4>IoT</h4></li>
<li class="sublink"><a href="lews.html">LEWS</a></li>
</ul>
</div>
<div id="page-content-wrapper">
<div class="page-content inset">
<div class="content-header"><h1 id="title">Publications</h1></div>
<div id="content"><p>
I've written some papers on both Machine-Learning and Cognition and considered listing them here:<br>
</p><ul>
<li><a href="https://drive.google.com/file/d/1zQfGqV2CPUqYc49CJmsUm5El6_zEhbqp/view">A Comparison of Class Imbalance Techniques for Real-World Landslide Predictions</a> (To appear in IEEE), 2017</li>
Landslides are a rare-phenomena. In training classifiers. if the "Occurred" instances are comparatively very low than "Not Occurred" instances then the classifier maybe biased towards the majority class and predict almost all instances as belonging to the majority class. On top of that, accuracy may be a misleading performance measure to evaluate the classifier in such cases. In this paper, we compare several minority class oversampling techniques like SMOTE, SMOTE-IPF, and Random Oversampling as well as use AUC (Area under the ROC Curve) and TP (True Positive) Rate as the performace metric on a dataset along Mandi-Manali Highway of India.<br> <br>
<li><a href="https://drive.google.com/file/d/1UVte1KR1p6to8VmdCD1k7zrfPpsVJayj/view">Hourly Scale Prediction of Landslides using Traditional Machine Learning Techniques</a> (To appear in Indian Landslide Congress), 2017</li>
In this paper, we try to overcome the daily-scale prediction of landslidess limitation and focus on an hourly-scale prediction of landslides by monitoring several soil and weather properties from a landslide site at Kamand, Himachal Pradesh. Data about temperature, humidity, rain, atmospheric pressure, light intensity, soil moisture, soil pressure, and soil movement are collected every 10-12 minutes from a landslide location on the Indian Institute of Technology Mandi campus at Kamand, Himachal Pradesh over a 10-day period in August 2017. We present preliminary results of the traditional machine-learning models like Random Forests, Support Vector Machines, C4.5 Decision Trees and Naive Bayes algorithms.<br> <br>
<li>Application of Reinforcement-learing Models to Understand Human Decision-Making in Interactive Landslide Simulator Tool (Under Preparations), 2017</li>
The main goal of this paper was to account for human performance decisions in the ILS tool via computational cognitive models and investigate the underlying cognitive processes involved. Specifically, we considered four different reinforcement-learning models (Expectancy Valence (EV), Prospect Valence (PV), Expectancy Valence Prospect Utility (EV-PU), and Propect Valence with a revised Utility function (PVL2) and evaluated their ability in capturing human decisions in the ILS tool in this work.<br>
<li>Patent: 201711045337 dated 18.12.2017 filed at the Indian Patent Office for proposing the Cheapest Monitoring and Warning System for Landslides</li>
<br>
</ul><p></p></div>
</div>
</div>
</div></body></html>