Libraries are a great repository of information and data that encapsulate community literacy, engagement, and health. However, this data is scattered across disparate sources. Our dashboard will help Anchorage Public Library collate this data and transform it into readable and usable information that can be used to drive insightful business decisions and highlight their achievements. The images and visualizations of our project will explain where and how the library’s resources are being used and how to effectively alter them if required. Our results will also recommend areas of improvement for APL along with the suggested formats to collect future information.

Social media companies are notoriously known for storing immense amounts of users’ personal data. While many social media sites do provide an option to view and download your data, the data is often given in raw formats, which can be overwhelming or hard to understand for the average user. Our project aims to analyze users’ downloaded data, and output it as data visualizations that are much more informative and intuitive. We hope that Aware will help people become more aware about what data is being stored about them and become incentivized to be more cautious about their online behavior.

Japanese seed beads are identified by the color code assigned to each product. Three primary manufacturers assign their own alphanumeric codes, while a wholesaler combined all three identification systems into a new system. This creates confusion among consumers because these systems are not well understood and are not yet cross-referenced. My ontology is the foundation of a knowledge base that will aid in more accurate identification of colors between manufacturer numbers and the corresponding product in the wholesaler’s system. This resource will facilitate accurate color-matching and increase awareness of the variety of seed beads available.

Doctors and parents in clinical trials face significant difficulties in tracking babies’ cries. The current method, handwritten crying logs, is inaccurate and offers little insight for clinical personnel. This project aims to use machine learning algorithms to more accurately identify and track cries. Hours of crying data were gathered and transformed into t-Distributed Stochastic Neighbor Embedding, Support Vector Machine, and k-Nearest Neighbor algorithms, visually grouping audio clips by sound similarity. An audio debugger determines the accuracy of the groupings. With continued training and testing, the algorithms will be able to engage binary classification of cries.

To help the Xbox team determine which technology performs better at detecting unsafe images, our project builds a comparison system that evaluates the performance of two image detection algorithms. The key results include: 1) a system that takes user images and generates similarity scores automatically; 2) an analysis pipeline that analyzes the data in Python and is live-connected with Power BI to show the key metrics. With our pipelines, the Xbox team can easily plug in images and see the comparison results on the dashboards. It eventually improves user experience by helping moderators remove the bad content effectively.

Our team set out to understand barriers to COVID-19 data publishing across the United States. Working with our sponsor, Sarah Schacht, we investigated how governments process and normalize a high volume of data as well as publish information to a variety of stakeholders. We conducted an ecosystem scan of all 50 states and key informant interviews to understand public health department needs. Using our findings, we produced reports highlighting shortcomings and obstacles in COVID-19 data reporting and provided recommendations. The report will hopefully be used to remedy existing problems and prepare government bodies to manage data in the future.

While the COVID-19 pandemic spared no region in the world and left in its wake an unprecedented public health and economic crisis, it has exposed and exacerbated pre-existing socioeconomic inequalities both globally and domestically. Using publicly available datasets, we have built a model that identifies neighborhoods in Miami that are most vulnerable to post-pandemic recession. This will enable government officials and policymakers to make data-informed decisions to spur the revival of the city’s economy.

During the COVID-19 pandemic, more activities have moved online but not everyone has access to reliable internet or the skills to do activities like school and work from home especially in rural communities. By analyzing data that has been collected by Digital Navigators using a set of intake forms, this project attempts to assess the lack of access and skills to reliable internet and devices, and provide necessary resources to rural communities in Appalachia. The results focused on the availability of the internet, access to the internet, especially in low-income households, and interest in employment.

Data-Driven Waste Management Solutions with trashcan pick-up frequencies are determined primarily by empirical observation. Seattle Public Utilities (SPU) sought a data-driven solution to better estimate waste levels and thus optimize waste bin placement and service scheduling. Our team detailed shortcomings of organization data collection and infrastructure, collaborated with stakeholders to create an appropriate data collection system, and created an optimization model to minimize costs associated with each collection route. Our efforts intend to reduce SPU’s expenses, support a cleaner city, and bolster SPU’s future information management practices to allow further opportunity. 

According to AP-NORC, 53% of Americans know nothing or very little about how money works in politics. After the Citizens United ruling allowed more money in politics, understanding campaign finance became more important than ever. Funds to Votes helps by displaying this data in a user-friendly and beginner researcher-friendly manner, allowing voters to make more informed decisions. Finally, in addition to helping ordinary citizens research their own representatives, our product is designed to aid endorsement groups in their research. This helps them decide who to endorse, which enables voters to select candidates according to their values.