1<!DOCTYPE html> 2<html lang="en-US"> 3 <head> 4 <meta charset="UTF-8"> 5 <meta http-equiv="X-UA-Compatible" content="IE=edge"> 6 <meta name="viewport" content="width=device-width, initial-scale=1"> 7 8<!-- Begin Jekyll SEO tag v2.8.0 --> 9<title>Portfolio | Palak Agarwal</title> 10<meta name="generator" content="Jekyll v3.10.0" /> 11<meta property="og:title" content="Portfolio" /> 12<meta property="og:locale" content="en_US" /> 13<meta name="description" content="Geospatial Data Scientist University of Pennsylvania Master of Landscape Architecture Master of Urban Spatial Analytics Landscape and urban designer bitten by the data bug. View My LinkedIn Profile" /> 14<meta property="og:description" content="Geospatial Data Scientist University of Pennsylvania Master of Landscape Architecture Master of Urban Spatial Analytics Landscape and urban designer bitten by the data bug. View My LinkedIn Profile" /> 15<link rel="canonical" href="https://palakagr.github.io/" /> 16<meta property="og:url" content="https://palakagr.github.io/" /> 17<meta property="og:site_name" content="Palak Agarwal" /> 18<meta property="og:type" content="website" /> 19<meta name="twitter:card" content="summary" /> 20<meta property="twitter:title" content="Portfolio" />
21<script type="application/ld+json"> 22{"@context":"https://schema.org","@type":"WebSite","description":"Geospatial Data Scientist University of Pennsylvania Master of Landscape Architecture Master of Urban Spatial Analytics Landscape and urban designer bitten by the data bug. View My LinkedIn Profile","headline":"Portfolio","name":"Palak Agarwal","publisher":{"@type":"Organization","logo":{"@type":"ImageObject","url":"https://palakagr.github.io/images/PalakAgarwal_Headshot3.jpg?raw=true"}},"url":"https://palakagr.github.io/"}</script>
22 23<!-- End Jekyll SEO tag --> 24 25 <link rel="stylesheet" href="/assets/css/style.css?v=312384abcea0c4f50a8b76afd899753c655a7fdc"> 26 <!--[if lt IE 9]> 27
27<script src="https://cdnjs.cloudflare.com/ajax/libs/html5shiv/3.7.3/html5shiv.min.js"></script>
27 28 <![endif]--> 29 </head> 30 <body> 31 <div class="wrapper"> 32 <header> 33 <h1 style="font-size:20px"><a href="https://palakagr.github.io/">Palak Agarwal</a></h1> 34 35 36 <img src="/images/PalakAgarwal_Headshot3.jpg?raw=true" alt="Logo" width="150" height="200"/> 37 38 39 <p style="font-size:14px">Geospatial Data Scientist <br> University of Pennsylvania <br> Master of Landscape Architecture <br> Master of Urban Spatial Analytics <br><br> Landscape and urban designer bitten by the data bug. <br><br> <a href="https://www.linkedin.com/in/palak-agarwal-025a68191/">View My LinkedIn Profile</a> 40</p> 41 42 43 44 45 <p class="view"><a href="https://github.com/palakagr">View My GitHub Profile</a></p> 46 47 48 49 </header> 50 <section> 51 52 <h1 id="portfolio">Portfolio</h1> 53 54<h2 id="projects">Projects</h2> 55<hr /> 56 57<h3 id="canopy-view---saving-tomorrows-tree-canopy-today"><a href="https://htmlpreview.github.io/?https://github.com/palakagr/tree-canopy-loss/blob/main/Tree_Canopy_Loss.html">Canopy View - Saving tomorrows tree canopy today</a></h3> 58<p>Tree canopy is defined as an area of land which, viewed from a birdâs eye view, is covered by trees. Trees offer essential benefits to cities, including mitigating the urban heat island effect, absorbing stormwater runoff, a habitat for wildlife, aesthetics, and recreation. A 2020 study published in The Lancet found that if Philadelphia reaches 30% tree canopy in all of its neighborhoods, the city would see 403 fewer premature deaths per year. Most of the tree canopy loss occurred in historically disenfranchised communities. In response, the City of Philadelphia has set essential milestones for conserving and increasing the current tree canopy in the city. In this analysis, we model tree canopy loss in Philadelphia and identify risk factors. We additionally assess the cityâs progress on the goals and how this varies across neighborhoods.</p> 59 60<p><img src="images/TreeCanopy.jpg?raw=true" /></p> 61 62<p><a href="https://palakagr.github.io/MUSAPracticum/">Web Application</a> <br /> 63<a href="https://github.com/palakagr/tree-canopy-loss">Data repository</a></p> 64 65<hr /> 66<h3 id="spark-safe---a-wildfire-vulnerability-indicator"><a href="https://htmlpreview.github.io/?https://github.com/palakagr/SparkSafe/blob/main/FinalProject.html">Spark Safe - A wildfire vulnerability indicator</a></h3> 67<p>California has more wildfires than any other U.S. state. In 2020, California saw one of the worst wildfire seasons in recorded history, and climate change could lead to even more wildfire destruction in the future (Yan et al., 2020; Miller et al., 2020). To support Cal Fire and the state of California in their important work protecting California from fire, we developed a mobile application called SparkSafe. Using the SparkSafe app, homeowners in 13 Northern California counties can enter their address and immediately learn their wildfire risk score. We used data from 2010-2017 to predict fires for 2018-2019 and then checked for its accuracy.</p> 68 69<p><img src="images/SparkSafe-combine.jpg?raw=true" /></p> 70 71<p><a href="https://www.youtube.com/watch?v=gXrzPI28f2Q&feature=youtu.be">Marketing Sales pitch</a> <br /> 72<a href="https://github.com/palakagr/SparkSafe">Data repository</a></p> 73 74<hr /> 75<h3 id="detecting-human-settlements-lacking-electricity-with-remote-sensing"><a href="/pdf/MUSA_650_Final_Project_Report.pdf">Detecting Human Settlements Lacking Electricity with Remote Sensing</a></h3> 76<p>Given the rate at which urbanization is occurring around the world, it is vital that new human developments have access to basic amenities such as sanitation, water, and electricity. Although these amenities are easily provided in developed countries, developing countries face enormous challenges to offering these basic resources to their citizens. In this project, we will be using satellite imagery to detect human settlements that have no electricity in order to identify where resources are needed most.</p> 77 78<p><img src="images/RemoteSensing.jpg?raw=true" /></p> 79 80<p><a href="https://colab.research.google.com/drive/1zL6M9BxGVBS1Mut0REusrz4e7Fe7EbLy">Colab script 1</a> <br /> 81<a href="https://colab.research.google.com/drive/1z939ZfD_NyeAGO7dFDdtMIfN7Rb7y7SS">Colab script 2</a></p> 82 83<hr /> 84<h3 id="noaa-tide-study">NOAA Tide Study</h3> 85<p>With climate change and the current climate scenario, understanding flooding and sea level rise patterns is of utmost important. So, I created a dashboard which visualizes and analyzes sea level rise for 16 cities across the country using historical data to predict future trends. Through understanding the technical aspects of NOAA data I learned how to visually express synthesized data into formats which are easily digestible to a variety of audiences. The disparate formats of NOAA data make understanding potential impacts of sea-level-rise and strengthening storm surges difficult for those not familiar with this type of data. I believe that Increasing the legibility of complex datasets like these help by providing a larger outreach and which in turn can help local citizens get involved in supporting policies that address it.</p> 86 87<p><img src="images/NOAA.jpg?raw=true" /></p> 88 89<p>The below links will take you to application pages which are interactive.</p> 90 91<p><a href="https://mybinder.org/v2/gh/palakagr/NOAA-Tide-Study/HEAD?urlpath=%2Fpanel%2FSeaLevelRise-app"><img src="https://mybinder.org/badge_logo.svg" alt="Binder" /></a> 92<a href="https://mybinder.org/v2/gh/palakagr/NOAA-Tide-Study/HEAD?urlpath=%2Fpanel%2FSeaLevelRise-app2"><img src="https://mybinder.org/badge_logo.svg" alt="Binder" /></a> 93<a href="https://mybinder.org/v2/gh/palakagr/NOAA-Tide-Study/HEAD?urlpath=%2Fpanel%2FSeaLevelRise-app3"><img src="https://mybinder.org/badge_logo.svg" alt="Binder" /></a></p> 94 95<hr /> 96<h3 id="predicting-home-values-for-boston-using-r"><a href="https://htmlpreview.github.io/?https://github.com/palakagr/Predicting-Home-Values/blob/main/MUSA508_Assignment2_Leah_Palak.html">Predicting Home Values for Boston using R</a></h3> 97<p>Zillow publishes âZestimateâ valuations for homes across the U.S. using an algorithm that draws on data from county and tax assessor records, multiple listing services, brokerages, and homeowner submissions. This project sought to improve Zillowâs housing market predictions for Miami and Miami Beach. We conceptualized our model using the hedonic model, which predicts home prices by summing the value of its constituent parts. Our model includes three types of variables: internal
97characteristics of houses, amenities/public services, and spatial structures.</p> 98 99<p><img src="images/Home2.JPG?raw=true" /></p> 100 101<p><a href="https://github.com/palakagr/Predicting-Home-Values">Data repository</a></p> 102 103<hr /> 104<h3 id="predicting-home-values-for-philadelphia-using-python">Predicting Home Values for Philadelphia using Python</h3> 105<p>Zillow publishes âZestimateâ valuations for homes across the U.S. using an algorithm that draws on data from county and tax assessor records, multiple listing services, brokerages, and homeowner submissions. This project sought to improve Zillowâs housing market predictions for Miami and Miami Beach. We conceptualized our model using the hedonic model, which predicts home prices by summing the value of its constituent parts. Our model includes three types of variables: internal characteristics of houses, amenities/public services, and spatial structures.</p> 106 107<p><a href="https://mybinder.org/v2/gh/palakagr/PredictiveModeling-HousePrices/HEAD?filepath=PredictiveModeling-Housing-Prices.ipynb"><img src="https://mybinder.org/badge_logo.svg" alt="Binder" /></a></p> 108 109<p><img src="images/predictions2.jpg?raw=true" /></p> 110 111<hr /> 112<h3 id="space-time-prediction-of-bike-share-demand---san-francisco"><a href="https://htmlpreview.github.io/?https://github.com/palakagr/Space-Time-Prediction-of-Bike-Share-Demand---San-Francisco-/blob/main/Space-Time%20Prediction.html">Space-Time Prediction of Bike Share Demand - San Francisco</a></h3> 113<p>In 2013 the San Francisco Municipal Transportation Agency partnered with public agencies to plan and implement the original Bay Area Bike Share pilot project and are now leading to work with the private sector partner to expand the system. One of the big operational challenges of bike share systems is âre-balancingâ - getting bikes to stations that are anticipated to have demand but lack bikes. Figuring out how to do this is one of the keys to operating a successful system. With the introduction of the stationless bikeshare system the problem of re-balancing can be eliminated by a certain percentage. But it creates new problems as the bikes are further spread out and its harder to make sure that the bike is picked up.</p> 114 115<p><img src="images/Bikeshare5.jpg?raw=true" /></p> 116 117<p><a href="https://github.com/palakagr/Space-Time-Prediction-of-Bike-Share-Demand---San-Francisco-">Data repository</a></p> 118 119<hr /> 120<h3 id="improving-hcd-tax-credit-programs-efficiency"><a href="https://htmlpreview.github.io/?https://github.com/palakagr/Improving-efficiency-of-HCD-tax-credit-program/blob/main/HosuingSubsidy_Palak_m.html">Improving HCD Tax Credit Programâs efficiency</a></h3> 121<p>The Department of Housing and Community Development (HCD) in Emil City seeks to launch a targeted campaign to encourage homeowners to take advantage of a $5,000 tax credit for home repairs. Typically, only 11% of the eligible homeowners they contact take the credit. This analysis attempts to improve the efficiency of HCDâs outreach effort, minimizing outreach to homeowners who are unlikely to take the credit while maximizing outreach to homeowners who are likely to take the credit. Using a random sample of HCD records, a logistic regression is developed to predict the probability that an eligible homeowner will take the home repair tax credit.</p> 122 123<p><img src="images/HCD2.jpg?raw=true" /></p> 124 125<p><a href="https://github.com/palakagr/Improving-efficiency-of-HCD-tax-credit-program">Data repository</a></p> 126 127<hr /> 128<h3 id="risk-modeling---predictive-policing"><a href="https://htmlpreview.github.io/?https://github.com/palakagr/Risk-Modeling---Predicitve-Policing/blob/main/MUSA508_Assignment3_Palak.html">Risk Modeling - Predictive Policing</a></h3> 129<p>One of the main aims of policing making is to make smart and efficient decisions in terms of resource and capital allocation which benefits large populations. It is always a comparison between initial investment to long trend usage and benefits. Policing has driven public-sector machine learning because law enforcement has significant planning and resource allocation questions. Here we look at the intersection of public sector planning and data science to ensure that the supply of a limited resource matches the demand for those resources. There are many traditional methods that exist but none have been proven effective.</p> 130 131<p><img src="images/Policing2.jpg?raw=true" /></p> 132 133<p><a href="https://github.com/palakagr/Risk-Modeling---Predicitve-Policing">Data repository</a></p> 134 135<hr /> 136<h3 id="urban-growth-patterns-around-international-borders">Urban Growth patterns around International Borders</h3> 137<p>When it comes to territorial disputes across the globe, the list is long and ever-changing. There are now more than 150 disputes under way that involve territory, mostly in Africa, Asia, and the Pacific region, but also in Europe and America. Many internal and external policies have contributed to this increase. The factors can be categorized
137in the large concepts of economic development and social development. Some cities grew as a result of increased trade between the countries while others grew because of other territorial disputes. Here I use Earth Engine and ArcGIS Python to analyze development trends around 7 borders.</p> 138 139<p><img src="images/1700.png?raw=true" /></p> 140 141<p><a href="/pdf/UrbanGrowthAroundContestedBorders.pdf">Final Report</a></p> 142 143<hr /> 144<h3 id="urban-street-networks">Urban Street networks</h3> 145<p>In this project, I used osmnx to analyze the crash incidence in Center City, Philadelphia. Crash data was found on PenDOT website that saves and calculates number of crashes each day. From the road networks, it was easy to identify different intersections which then were used to geolocate the crashes reported.</p> 146 147<p><a href="https://mybinder.org/v2/gh/palakagr/Urban_Street_network/HEAD?filepath=Urban_street_networks.ipynb"><img src="https://mybinder.org/badge_logo.svg" alt="Binder" /></a></p> 148 149<p><img src="images/urban_street.jpg?raw=true" /></p> 150 151<hr /> 152<h3 id="yelp-review-analysis">Yelp review analysis</h3> 153<p>In this project, I explored restaurant review data available through the Yelp Dataset Challenge for Cleveland. The dataset includes Yelp data for user reviews and business information for 10 metropolitan areas. I ran sentiment analysis on the reviews to check the polarity and subjectivity. Because Yelp reviews include the number of stars given by the user, the Yelp data set provides a unique opportunity to test how well our sentiment analysis works by comparing the number of stars to the polarity of reviews. Then analyzed the frequencies of different sets of words in Yelp reviews can offer insight into urban culture. Specifically, test the correlation between household income (using census data) and different types of restaurants, extracted from review data.</p> 154 155<p><a href="https://mybinder.org/v2/gh/palakagr/Yelp_Reviews/HEAD?filepath=Yelp_Reviews.ipynb"><img src="https://mybinder.org/badge_logo.svg" alt="Binder" /></a></p> 156 157<p><img src="images/yelp.jpg?raw=true" /></p> 158 159<hr /> 160<h2 id="design-projects">Design Projects</h2> 161<p>Design and analytics for me is a complex matrix composed of tangible and intangible features. I believe that enactable policies should be developed and employed across the country, reflective of the myriad spatial, social, cultural, political, and morphological agents that constitute place. Understanding the relationship between the multiplicity of actors is the key to solving many of the urban problems we face today. To find solutions that work for today and tomorrow, it is important to grasp patterns that they create, and data can help us do that.</p> 162 163<p><img src="images/Design2.jpg?raw=true" /></p> 164 165<p><a href="/pdf/PalakAgarwal_Resume.pdf">Resume</a> <br /> 166<a href="/pdf/Agarwal_Palak_Portfolio.pdf">Design Portfolio</a></p> 167 168<hr /> 169<h2 id="research-projects">Research Projects</h2> 170 171<h3 id="field-notes-toward-an-internationalist-green-new-deal"><a href="https://internationalistgreennewdeal.org/">Field Notes toward an Internationalist Green New Deal</a></h3> 172 173<p>Field Notes toward an Internationalist Green New Deal is a project of the <a href="https://mcharg.upenn.edu/">McHarg Center</a> under the direction of Billy Fleming. Employing a variety of digital media, mapping, storytelling, and other methods, this project asks a series of critical questions about the international and justice dimensions of the Green New Deal. How might the buildout of clean energy infrastructure in the Global North reproduce exploitative relations to the Global South? How might the Green New Dealers seek to ameliorate concerns that their vision represents yet another a form of green imperialism or green capitalism? How might designers think differently about where and how they participate in the discourse around an energy transition, a just transition, and a Green New Deal?</p> 174 175<p><img src="images/image.png?raw=true" /></p> 176 177<hr /> 178<h3 id="remote-sensing-and-tangible-modeling">Remote Sensing and Tangible modeling</h3> 179 180<p>In my research position at the Environmental Modeling Lab at the University of Pennsylvania I work intimately with satellite imagery to find spectral signatures for different kinds of systems. We utilize machine learning algorithms to train and test different images across different seasons and years. Also we work on pansharpening multi-spectral imagery and also collecting imagery from the field using drones.</p> 181 182<p><img src="images/remote.jpg?raw=true" /></p> 183<p style="font-size:6px">Photo credit : Jayson Latady</p> 184 185<hr /> 186<h2 id="skills-and-technical-expertise">Skills and technical expertise</h2> 187 188<ul> 189 <li><strong>Geospatial Analytics:</strong> ESRI GIS Software Suite (raster and vector), ArcPy, Google Earth Engine, Environmental Modeling, QGIS</li> 190 <li><strong>Programming Languages:</strong> R, Python, JavaScript, HTML, C++, Git, OpenStreetMaps</li> 191 <li><strong>Statistical Analysis:</strong> Machine learning (supervised and unsupervised), Predictive Modeling in R, Spatial Statistics</li> 192 <li><strong>Data Visualization:</strong> HTML/CSS, Mapbox, Carto, ArcGIS API, Adobe Suite, Seaborn, Matplotlib, Altair (Python)</li> 193 <li><strong>Web Development:</strong> Carto, JavaScript, Python for Web Development (panel, folium, flask)</li> 194 <li><strong>Satellite Imagery/ Remote Sensing:</strong> ArcGIS Spatial Analytics Toolbox, Python image processing</li> 195 <li><strong>Landscape Design:</strong> Land Use and Environmental Planning, Land Preservation, Urban Finance, Hydrology, Designing skills</li> 196 <li><strong>Environmental Science:</strong> Environmental restoration, Climate-change concepts, Ecological topics</li> 197 <li><strong>3D Modeling softwares:</strong> Rhino, Sketchup, Lumion</li> 198</ul> 199 200<hr /> 201<p style="font-size:6px">Page template forked from <a href="https://github.com/evanca/quick-portfolio">evanca</a></p> 202<!-- Remove above link if you don't want to attibute --> 203 204 205 </section> 206 <footer> 207 208 <p><small>Hosted on GitHub Pages — Theme by <a href="https://github.com/orderedlist">orderedlist</a></small></p> 209 </footer> 210 </div> 211
211<script src="/assets/js/scale.fix.js"></script>
211 212 213
213<script>
vendor: 364 bytes, lines 213-218
213 214 (function(i,s,o,g,r,a,m){i['GoogleAnalyticsObject']=r;i[r]=i[r]||function(){ 215 (i[r].q=i[r].q||[]).push(arguments)},i[r].l=1*new Date();a=s.createElement(o), 216 m=s.getElementsByTagName(o)[0];a.async=1;a.src=g;m.parentNode.insertBefore(a,m) 217 })(window,document,'script','https://www.google-analytics.com/analytics.js','ga'); 218 ga('create', '
218UA-000000-0
vendor: 46 bytes, lines 218-220
218', 'auto'); 219 ga('send', 'pageview'); 220
220</script>
220 221 222 </body> 223</html>
Line numbers count LF bytes from the start of the resource, as the search results do. Vendor segments are library code the classifier recognised; they are stored but not indexed. Bytes are shown as Latin1 characters, one per byte.