Research Talks

Session I: March 24 at 12:00PM

Skin Cancer Prevention: A MyVoice and IMPACT Melanoma Project

Arianna Strome, Institute for Healthcare Policy and Innovation, University of Michigan
Kelsey Herbert, Institute for Healthcare Policy and Innovation, University of Michigan
Tammy Chang, Institute for Healthcare Policy and Innovation, University of Michigan

Purpose: To transform our community’s approach to skin cancer prevention by making sunscreen accessible to all in Washtenaw County and on the University of Michigan campus.

Background: Skin cancer is the most common type of cancer in the United States with incidence continuing to rise. The leading risk factor for skin cancer is sun exposure in adolescence. Given this, sun protection in young adults is an effective and critical way to reduce cancer occurrence. In our nationwide, text message-based study of youth aged 14-24 with MyVoice (, we discovered youth understand both the short- and long-term risks of sun exposure and believe that sun protection is important. Nearly all youth reported sunscreen use, though nearly all also reported experiencing numerous sunburns. Our findings suggest there is a discrepancy between knowledge and real-life implementation of sun protection due to access and convenience. (Strome & Herbert, JAMA Network Open, 2021)

General Information: Our project uses these findings to improve the well-being of our local community by increasing access to sun protection. We have partnered with the non-profit IMPACT Melanoma to install sunscreen dispensers in high sun-exposure areas such as sporting venues. These dispensers provide sun protection and encourage dialogue about the importance of applying sunscreen. Our goal is to encourage dialogue about the importance of applying sunscreen and to create a culture of empowerment around sun protection that is accessible to all community members. We are also working towards obtaining the Skin Smart Campus designation from the National Council on Skin Cancer Prevention by eliminating access to indoor tanning units on campus.

Future endeavors: Our goal is to remove barriers to sun protection through equitable and community-based solutions such as dedicated sunscreen dispensers located in public high sun areas throughout the state of Michigan.

Experience of stigma and its relationship to identification with the neurodiversity model for Indian parents of children with autism spectrum disorder

Sahita Manda, School of Public Health, University of Michigan
Elizabeth Buvinger, PhD, Department of Psychology, University of Michigan
Shichi Dhar, College of Literature, Science, and the Arts, University of Michigan
Harika Veldanda, College of Literature, Science, and the Arts, University of Michigan

It is widely recognized that individuals with autism spectrum disorder (ASD) and their families continue to face extensive stigma and that much of the current research on ASD is deficit-focused. Diversity and inclusion perspectives are emerging, but there is less of a focus on how stigma affects the adoption of these approaches. In collaboration with the University of Michigan Department of Psychology and the national Indian organization Action for Autism, our research aims to understand the experience of stigma and its relationship to identification with the neurodiversity model for Indian parents of children with ASD. The study was carried out by administering online surveys through the platform of Qualtrics to Indian parents residing in India (N=56). This study explores the extent to which Asian value adherence, child functioning, and perceived ASD stigma contribute to parental alignment with the neurodiversity model. It also investigates the ways in which alignment with the model affects parental stress, isolation from family and friends, parenting goals, identification of child’s strengths, and positive perceptions about raising a child with ASD. Preliminary findings demonstrate statistically significant correlations between a child’s ASD behaviors, perceived ASD stigma, parental stress, and isolation from family and friends. A more complex mediation model of the effects of neurodiversity alignment on these variables will be presented and will have implications for the adoption of strength-based practices and the reduction of stigma associated with ASD within different cultural contexts.

Pop-up Safety Town: Leveling the Playing Field for Children in High-need Communities

Andrew Hashikawa, MD, MS, Departments of Emergency Medicine and Pediatrics, Michigan Medicine
Erin Kim, BS, University of Michigan Medical School

Background: Pediatric unintentional injuries are among the most under-recognized public health issues in the United States. An “upstream” approach involves providing injury prevention education to preschool children. However, the existing intervention (Safety Town) was not accessible for families living in resource-poor neighborhoods. To address this gap, we created a mobile, budget-friendly intervention (Pop-up Safety Town) and partnered with federal Head Start (HS) Preschool Centers to bring injury prevention programming to families.

Methods: Michigan Medicine and community health and safety expert volunteers worked with HS children (ages 3-5 years) and their parents. Our events used “pop-up” buildings to create miniature towns, easily set up, transported, and reused. Our core programming included helmet, firearm, passenger, water, animal, electrical, medication, dental and pedestrian safety stations. Participants rotated through stations every 15 minutes and received meals, bike helmets, and books as incentives. Surveys were obtained for feedback.

Results: Over 1.5 years, five events were held with 135 families and 235 children; >90% of parents reported sessions were “extremely or very useful,”; Our pilot demonstrated 1) HS centers are excellent locations for injury prevention education in resource-poor neighborhoods; 2) most families (>95%) had never before received injury prevention education programming; 3) many families lacked resources (~50% of children did not own bike helmets); 4) parents continued to discuss topics with their children several months after the event.

Future: In 2021, we received additional funding from the Michigan Health Endowment Fund and the Michigan AAA Auto Club Foundation. Our goals are to expand our partnership to the Detroit Mary Grove Center and expand to other remote sites, including the American Indian and the Migrant/Seasonal HS. We aim to spearhead a trial using a virtual format incorporating interactive educational videos emulating the in-person safety stations and comparing outcomes (virtual vs. in-person sessions) to help mold future sessions.

Cancer tracking and management in Ghana using Data: A Zurak Cancer Foundation Story

Suleman Mahmoud Junior, School of Information Technology, University of Cincinnati, Zurak Cancer Foundation, Ghana
Abdul Samed Zurak, Zurak Cancer Foundation, Ghana

Zurak Cancer Foundation is a policy-oriented and service delivery youth-run non-governmental organization (NGO) providing innovative solutions to cancer prevalence in Ghana through education, awareness program, skills acquisition for medical professionals and screening programs in a slum, rural and hard to reach communities in Ghana.

Every year, about 311,000 women die from cervical cancer, and ninety percent of these cases occur in low- and middle-income countries. Treatment of cancer in Ghana is not only expensive but makes it difficult to manage and track health status of Suspected cases due to the data deficiency in the healthcare sector. Zurak Cancer foundation has improved this by collecting data, using machine learning algorithms to analyze for real time results and partnering with stake holders to bring effective solutions to the beneficiaries of the project.

The focus areas and places of impact are in the three (3) cities of Ghana namely Accra, Kumasi, and Tamale and in partnership with Johnson and Johnson, Roche Products, Visualize Ghana. The survey method was used in collecting and analyzing data, Data used to support these projects were collected through Market invasions, Awareness programs, trainings, and online campaigns. Data collection mediums used by volunteers to collect data were google forms for real time analysis and a mobile app designed by the innovation team. We analyzed the data using machine learning algorithms to identify and detect potential cancer cases.

Our Breast cancer and cervical cancer awareness programs are both online and on ground. For online, our targeted organic reach on our social media platforms are 34,751 people . We targeted 1,000 market women for our Market Invasion and together making a total 35,751 or people for the cancer awareness months. Women with suspected cases were enrolled in a patient management program to provide access to treatment and further diagnosis.

Session II: March 25 at 12:00PM

Revisiting the Relationship Between Clinical Mentor and Student Teacher Effectiveness

Emanuele Bardelli, University of Michigan
Matthew Ronfeldt, University of Michigan
Lauren Fisher, University of Michigan

Recent research shows what has long been suspected: pre-service teachers (PSTs) who complete their clinical placements with more effective clinical mentors (CMs) go on to become more effective teachers than their peers. Other research found that the characteristics of field placement schools (FPSs) where PSTs complete their clinical preparation are related to PSTs’ performance; PSTs perform better when their employment schools more closely match their FPSs. Our work grew out of a community-university partnership between our team and a state department of education with the goal of improving teacher preparation outcomes in the state. To do this, we replicate prior work examining the relationship between the general instructional effectiveness of PSTs and CMs with an extension to most programs in the state and consider specific domains of the teacher evaluation rubric. We examine heterogeneity in the relationship between CM and PST instructional effectiveness across various dimensions of FPSs (e.g. school size, grades served). Consistent with prior work, PST ORs are positively associated with CM ORs (b=0.037, p<0.001); however, they are negatively correlated with CM VAMs (b=-0.045, p<0.05). PST VAMs are positively correlated with CM VAMs (b=0.051, p<0.01). PST VAMs are not significantly related to CM ORs, however there is a positive trend.

We also find that PST effectiveness in specific teaching domains tends to be correlated with their CM’s effectiveness in the same domains, suggesting that the relationship between CM and PST instructional effectiveness may be causal. There is a significant positive correlation between PST and CM ratings in the instruction (b=0.035, p<0.01) and planning domains (b=0.039, p<0.001). Heterogeneity by FPS characteristic analyses are ongoing. Preliminary findings suggest that these effects are heterogeneous for FPS school level (e.g., elementary vs. secondary) and PST endorsement area (e.g., secondary math vs. ELA).

Genre within Data: A Rhetorical Approach to Public Good

Kelly L Wheeler, Joint Program English and Education, University of Michigan
Jacob Richman, Data Science, School of Literature, Science, and the Arts, University of Michigan

To make sense of large amounts of data, researchers are tasked with looking for and establishing patterns, but what are the material and embodied consequences of establishing patterns within that data? How does determining a pattern function to give meaning to researchers and communities and support social justice? This presentation uses rhetorical genre studies as both a theoretical lens and a methodology to understand patterns within the The Swastika Counter (TSC), which is a database comprised of 1339 online news stories that documents swastika appearances across the United States from January 2016-January 2020. In looking at the patterns created by the data within TSC, we offer up our own definition of a rhetorical genre: community response to swastika hate acts. Repetition of certain assemblages of discursive products and non-discursive actions of actors defines this particular genre, and those assemblages commonly perform the rhetorical moves of silence, speech, and support. How those assemblages perform those moves manifest themselves in both “ordinary” and “extraordinary” ways, and in establishing a baseline of “ordinary” moves in contrast to “extraordinary,” we have been able to press upon communities’ responses to make visible practices of power, inequity, and embodied harm when traumas like swastikas occur in a community. Using data to define community response to hate acts as a genre creates opportunities for scholars and communities to critically engage in how their own communities, both local and academic, respond to swastika hate acts. Critical reflection and conversation about community responses could then lead to better enactment of social justice within those communities and possible mitigation of harm to marginalized communities within those communities in the future.

The GRE in public health admissions: Using public data to influence institutional change

Jess Millar, Computational Medicine and Bioinformatics, University of Michigan

In the wake of COVID-19, there is an urgent need for a diverse public health work force to address problems presented or exacerbated by the global pandemic. Educational programs that create our work force both train and shape the makeup of access through graduate applications. The Graduate Record Exam (GRE) has a number of standing issues, with additional barriers created by the pandemic.

We trace the GRE waiver movement over several years using publicly available data from university websites, focusing on the gradual adoption in Council on Education for Public Health (CEPH) accredited programs and the rapid expansion of temporary waivers as a response to testing access. We document the GRE waiver status for all CEPH program concentrations / degrees periodically between October 2019 to October 2021, including whether waiver was temporary or permanent. Up to date data is made available to the public as a google spreadsheet for students to make application decisions and to be used as leverage by departments. GRE waiver use has expanded over 840% in 2 years, with 87% of CEPH programs offering at least one waiver. Permanent waivers for concentrations / degrees increased from 10% to 56%, with temporary waivers making up an additional 27%. However, the coverage within and between programs is more inconsistent.

By following and reporting the GRE waiver status for all CEPH degrees, it was possible to show the scope of adoption in the field and use that as leverage to influence buy in power to other institutions. As universities became more aware of other universities waiving the GRE, emails to join the list increased as well as overall adoption of waivers. Going forward, we need to consider gaps in waivers during the pandemic and how this data can be used to shape our future use of the GRE.

Developing course equity reports to understand and reduce inequity in University of Michigan classes

Nicholas T. Young, Center for Academic Innovation
Rebecca L. Matz, Center for Academic Innovation
Heather Rypkema, Center for Research on Learning and Teaching
Susan Cheng, Center for Research on Learning and Teaching
Holly Derry, Center for Academic Innovation
W. Carson Byrd, National Center for Institutional Diversity
Ben Koester, Department of Physics
Eric Bell, Department of Astronomy
Caitlin Hayward, Center for Academic Innovation

Grades play an important role in determining whether students perceive themselves as competent in a domain, can get into and stay in the program and major of their choice, keep their financial aid and scholarships, and ultimately graduate from the university. Yet, systematic differences in grades earned that are correlated to gender, race, socioeconomic class, and parental education lead to inequities in a host of outcomes such as which degrees students of diverse backgrounds choose to pursue. One focus for reducing inequity in higher education is examining inequities at the level of individual courses. To do so, we used the University of Michigan’s Learning Analytics Data Architecture to develop an automated report that analyzes individual university courses. Through this course equity report, instructors can visualize students’ backgrounds and trends in longitudinal grading patterns. In addition, these course equity reports provide concrete suggestions that instructors can consider for making their courses more equitable. In this presentation, we will describe how we created these reports, talk through specific examples of the visualizations contained within, and how members of the Michigan community are currently involved in the project and how others can get involved.

Session III: March 25 at 2:00PM

Creating Political Windows for Persons Experiencing Homelessness: Problems, Politics, Policies, and COVID-19 Vaccination

Gabriella Vanaken, University of Michigan Medical School
Payge Barnard, University of Michigan Medical School
Brent Williams, Department of Internal Medicine, Michigan Medicine

Persons experiencing homelessness (PEH) in Michigan were not included in the COVID-19 vaccination phasing schedule until March 8th, 2021, at which point they were placed in Phase 1B, which had begun two months earlier on January 11th, 2021. Despite equivalent living conditions and health risk factors that allowed others to qualify for earlier phasing, PEH were not placed in Phase 1A. In addition to congregate living, PEH are vulnerable for multiple reasons. They face psychosocial barriers resulting in physical ailments arising 15 to 20 years earlier than their housed counterparts. Throughout the pandemic, this population has been twice as likely to be hospitalized, in need of critical care, and die than housed individuals. Additionally, minority groups experience homelessness at rates 2 to 14 times that of white individuals. Racial disparities in homelessness are particularly prominent in Detroit, which has the highest rates of homelessness in the State of Michigan and the third highest population of black individuals in the U.S. The purpose of this study is to utilize the Policy Windows framework to analyze phasing policies of the COVID-19 vaccine for PEH in Washtenaw County and the City of Detroit, and to determine how windows of opportunity were recognized and utilized in order to vaccinate PEH. We also aim to determine advocacy efforts and their impact on the creation of political windows for COVID-19 vaccination of PEH. This study is a thematic analysis of data collected from interviews with the Michigan Department of Health and Human Services, Washtenaw County Health Department, Detroit Health Department, Neighborhood Service Organization, Packard Health, and Robert J. Delonis Center. Data was also gathered from publicly available online sources. Detroit’s problem, policy, and politics stream were in alignment and allowed for early vaccination of PEH as opposed to Washtenaw County, whose population of PEH suffered.

Compiling a Census Assessment Tool

Amy O’Hara, Massive Data Institute, McCourt School of Public Policy, Georgetown University
Ronald Prevost, Massive Data Institute, McCourt School of Public Policy, Georgetown University
Christopher Dick, Massive Data Institute, McCourt School of Public Policy, Georgetown University
Izzy Youngs, Massive Data Institute, McCourt School of Public Policy, Georgetown University
Claire Bowen, Center on Labor, Human Services, & Population, Technology & Data Science, Urban Institute

Problem: Accurate decennial census data are critical to making effective decisions at the national, state, and local levels. The Census Bureau’s population counts affect congressional redistricting, social program funding, school planning, and more. The importance of the decennial census is clear, but the 2020 Census faced unprecedented challenges with unknown effects on data quality, from the impact of the COVID-19 pandemic and related stay-at-home orders to new household compositions, displacement of college students, and natural disasters. These factors have led some data users to question whether the data are fit to use.

Tool: In response, the Massive Data Institute in collaboration with the Urban Institute created the 2020 Census County Assessment Tool, in order to assist data users in identifying deviations between expected counts and the released counts across population and housing indicators. The tool also offers contextual data for each county on factors which could have contributed to census collection issues. The tool compiles this information into a downloadable report and points users to additional local data sources or experts to seek more assistance.

Benefit/Impact: This tool was developed for planners, administrators, policymakers, teachers and students, reporters, researchers, and anyone interested in census data in order to assist in identifying strategies for utilizing census data appropriately. This tool particularly provides resources to data users interested in the accurate enumeration of race and ethnicity, pointing users to areas which may need additional attention for the 2030 census. This tool can assist users in intercensal planning, count review requests, population estimates challenges, funding allocation corrections, Get Out the Count investment strategies, and more.

‘Yorme, Yorme’: Modeling themes from public COVID-19 reports aired through Manila City government Twitter accounts

Joseph Marvin Imperial, Computer Science and Human Language Technology Lab, National University, Philippines

In order to formulate effective data-driven decisions for improvement of services, local government units (LGUs) and their constituents should implement a bottom-up approach where the primary source of data is from the ground or directly from the people. In this work, we propose an AI-based framework for computationally and automatically understanding tweets sent by the people of Manila City, Philippines directed to the official accounts of Mayor Isko Moreno and Manila PIO, the two primary accounts serving as sources of information for city-wide announcements and events. We collected over 7,405 public tweets and tweet replies sent within the Manila City area during the first two months of Enhanced Community Quarantine (ECQ) from March 1 to April 6, 2020. Our framework is composed of three subtasks derived from artificial intelligence methods which are Biterm Topic Modelling (BTM), Bigram Analysis, and a simple Named Entity Recognition (NER). Through BTM, we uncovered four major recurring themes present from the tweet reports sent by the netizens of Manila City: (a) repetitive expressions of needs such as relief goods and aid, (b) reports of mistreatment from police and barangay officials, (c) call for use of public transportation such as tricycles for use, and (d) call for suspension of classes. Applying Bigram Analysis exposed co-occurring terms frequently used to build context for the reports. Finally, using a simple RegEx-based NER allowed specific tracking of which barangay in Manila City is targeted in the reports. We emphasize that these findings and the proposed framework can be adapted and scaled by local government units to consider public opinion and call of the masses to improve overall people-centric services.

Censored Planet: An Internet-wide, Longitudinal Censorship Observatory

Ramakrishnan Sundara Raman, Computer Science and Engineering, University of Michigan
Roya Ensafi, Computer Science and Engineering, University of Michigan

Censored Planet is a global and longitudinal Internet censorship measurement platform that identifies and monitors Internet censorship in more than 200 countries. Censored Planet uses remote Internet measurement methods to measure the blocking of websites, removing the need for volunteers in different countries who may be at risk collecting censorship measurements. Since its launch in August 2018, censored Planet has been continuously collecting reachability data pertaining to 2000 popular and sensitive websites every week from more than 95,000 vantage points around the world. So far, Censored Planet has collected more than 45 billion measurement data points, running into several Terabytes of data, which is made available to the public through the Censored Planet website. More than 60 organizations in the Internet Freedom and technology community have requested access to Censored Planet data for their research or advocacy, and Censored Planet data has been crucial in understanding and responding to censorship, surveillance, and network interference events such as the large-scale HTTPS interception attack in Kazakhstan and the throttling of Twitter in Russia. Censored Planet and its findings on multiple censorship events have been covered by NPR, MIT Technology Review and the Financial Times. Censored Planet continues to produce open-source censorship measurement data to date, and we are working closely with organizations such as Google Jigsaw to help make the data more useful for researchers and practitioners.

Session IV: March 25 at 3:00PM

Drinking Water Data Learning Tools for Small Utilities

Rebecca Hardin, School for Environment and Sustainability, University of Michigan
Matt Vedrin, Civil and Environmental Engineering, University of Michigan
Lut Raskin, Civil and Environmental Engineering, University of Michigan
Ed Waisanen, School for Environment and Sustainability, University of Michigan
Kyle Powys Whyte, School for Environment and Sustainability, University of Michigan

This Public Interest Technology project, funded by New America Foundation, leverages local partnerships to co-create prototype digital modules, piloting them for more inclusive, dynamic, data-rich approaches to drinking water monitoring and treatment at regional scales, with inclusive educational reach. Water utilities across the state of Michigan report that they face severe challenges in replacing current systems management expertise and need to attract young learners to future water workforces. Thus, we propose using open-source, open-access technology to create interactive, responsively designed modules, then pilot, revise and adapt them for use in both professional development contexts and classrooms to incite interest in public utility careers and enhance data skills of existing workforces.

Preventing Flooding Through Data Visualization

Tess Eschebach, Department of Civil and Environmental Engineering
Jacquelyn Schmidt, Department of Civil and Environmental Engineering
Branko Kerkez, Department of Civil and Environmental Engineering

Flooding is the leading cause of natural disaster deaths in the world. As climate change makes extreme weather more common, cities are discovering that their stormwater infrastructure is no longer able to hold back the water. Improving stormwater infrastructure the traditional way, by building larger stormwater basins, storage tanks, and pipes, is too expensive for most cities to afford. Instead, water utilities around the world are investing in “smarter” stormwater systems — retrofitting existing networks with sensors, robotics, and machine-learning based algorithms, to create adaptable stormwater systems capable of handling much larger storms.

While the engineering has come a long way, water infrastructure UI/UX has not kept up. We present a review of existing stormwater management UI/UX design practices, focusing on the benefits and pitfalls of these interfaces. Clear data visualization is essential for data-driven decision making, and as new sources of data are introduced into this field, it is increasingly important to understand how to design good stormwater data interfaces. Taking cues from similar fields such as electrical grid management and seismic monitoring, we offer a set of guidelines for modernizing data visualization in stormwater management. The findings of this literature review will be tested in a real-world setting through our partnership with the Oakland County Water Commissioner’s Office. Water elevation interfaces are utilized within Oakland County to regulate lake levels, preventing water from flooding residential areas. Oakland County’s particular needs are explored through a user study conducted through interviews and digital correspondence. Guided by best practice, we are working with stormwater managers from Oakland County to develop custom features and application-specific dashboards to help better regulate water levels.

Using Mobile Phone Data to Evaluate Access to Essential Services Surrounding Disruptive Events

Tessa Swanson, Industrial & Operations Engineering
Seth Guikema, Industrial & Operations Engineering, Civil & Environmental Engineering

Natural hazards bring about change in access to essential services through closure of facilities, transportation system disruption, evacuation orders, power outages, and other barriers. Understanding changes in access to essential services following a disruption is critical to ensure just and equitable recovery and more resilient communities. Location based services data collected from cell phones offer new opportunities for identifying patterns of unique visits to facilities and subsequently, deviations from those patterns that may indicate unavailability of those facilities, i.e. functional closures. In this analysis, we introduce a data-driven approach to identify functional closures at the facility level by assessing patterns in unique user appearances over time from LBS data. We show the application of this approach applied to samples of supermarkets, schools, health care facilities, and home improvement stores in Southwest Florida leading up to and following the landfall of Hurricane Irma in 2017. We then extend this method to individual-level appearances at user homes and workplaces. We quantify anomalies in home and workplace appearances to identify periods of workplace functional closures, thus estimating “return-to-productivity” as a proxy for individual-level recovery times. This work promotes the adoption of LBS data for hazards and risk analysis research for understanding facilities’ and individuals’ response to disruptions and the factors contributing to those behaviors.

A Data-Driven Approach to site selection of Pandemic resilient Fair Price Shops (FPS) in an urban context

Siddharth Sekhar Singh, SETS Pvt Ltd

Indian Public Distribution System (PDS), the leading national food security program of India, has witnessed several technology-enabled performance improvement initiatives over the last decade. The primary motivation behind these initiatives is to reduce leakages and improve transparency.

The necessary foundation of the Public Distribution System depends upon four factors viz. availability, accessibility, acceptability, and affordability. Aadhaar enabled public distribution system (AePDS), and end-to-end computerization of the Targeted public distribution system (TPDS) improved the availability and acceptability of subsidized food grains. However, in most cases, proper selection of the site location of fair price shops(FPS) (also known as ration shops) is improper, which inconveniences the beneficiaries and the ration shop owners alike. For instance, approximately 60% of ration shops in Delhi were established before the year 1990. Many ration shops are no more accessible to trucks ferrying ration from food corporation of India(FCI) warehouse. It results in labor and transportation cost escalation for ration shop owners making it difficult for the ration shop owner to sustain the shop. Long queues are a distinctive feature of ration shops. During our field visits, we observed that the site locations for a large number of shops are not convenient for the beneficiaries who visit these ration shops. These fair price shops represent the food and civil supplies department’s presence in communities across the nation to improve the quality of life of these communities. A proper site location of these fair price shops matters to the convenience of residents as well. In this paper, we introduce a novel approach to the systematic site selection for pandemic resilient ration shops in an urban context. It takes as input a set of data about a metropolitan area along with COVID-19 containment zone data and generates an­ optimal configuration of two-dimensional locations for urban sites appropriate for fair price shops that an be used to delivery food grains during lockdowns. We achieve this goal by using data-driven optimization, entangling deep learning. The proposed approach can cost-efficiently generate a site location plan considering representative site selection criteria, including coverage, accessibility, and convenience.