Capstone Projects | Information School | University of Washington

Project Rainier

The City of Seattle provides effective information to Seattleites through their website – www.seattle.gov. Seattle IT struggles to understand user behavior and effectiveness of their website. They spend a lot of time creating content. However, they are unclear if this content is surfaced in ways that are comprehensible for users. Our project is an effort to develop a KPI-based framework that content creators can use to understand users and enhance the discoverability of content. Through our holistic and comprehensive framework, Seattle IT can regularly track performance and drive continuous improvement of the online experience that users and potential users have.

Safistation™ Chlorine Generator Dashboard

PATH, a global health non-profit organization, has created the SafiStation, a chlorine generator to be used in low-resource health centers as a means of creating life-saving disinfectant. Our capstone team used data from the station to improve the current data dashboard that PATH is using to monitor their dashboards by developing specific user interfaces and visualizations that are individualized for each customer segment. Using the improved dashboard, PATH and their future customers can visualize how their machines are operating as well as the returns on their investments.

Semantic Analysis

SEMANTIC ANALYSIS: The Smartsheet platform has large amounts of unstructured user feedback comments. So, the product owners at Smartsheet take a lot of time and effort to come up with actionable feedback. Our solution is a machine learning model which enables text classification of the comments by categorizing them into the following categories- Bugs/Performance, Formula, UX Integration, and UX Enhancement. The solution makes it easier to prioritize issues, delegate the requests to the concerned Product Owners, reduces the response time to user queries, and enhances the product features. Organization - Smartsheet Team Members - Ishita Shah Prithvi Shetty Vineet Kulkarni

Semantic Analysis - Text Classification of Unstructured Customer Comments

Smartsheet is a cloud-based work management platform. To continuously innovate and improve its product and services, it constantly gauges customer sentiment and feedback via Net Promoter Score surveys. However, due to large inflow of raw comments from these surveys, and no existing mechanism to aggregate the comments other than manual intervention, the useful data is underutilized. Our solution is to apply machine learning techniques to automatically categorize the comments into useful categories. This categorized data is then presented as a dashboard to internal stakeholders to gain deeper insights into Smartsheet features and services.

SmartInvest: Sentiment Driven Insights using NLP

It is a nightmare for fundamental securities analysts and portfolio managers to leaf through lengthy corporate filings and call transcripts to gain insights on which stock to invest in. Our project automates this process using NLP tools and generates sentiment/credibility index. These indexes are used to create an all-inclusive dashboard that displays a company’s performance, its overall ranking, ranking within its industry, current trends in the market and popular buzzwords in different industries. The portfolio managers can then use the dashboard to make smarter investment decisions. Do you also wanna make your investments more profitable? Try SmartInvest!

Spectrum

Spectrum is a melting pot of different sources and types of news, all on one website. We challenge individuals to read from multiple sources and categories and encourage users to reflect on their personal reading habits. We challenged ourselves to create a place where people can leave with a more well-rounded understanding of current events in the world we live in. In the end, we hope to impact everyone who comes across Spectrum and inspire individuals to be aware of various viewpoints and generate mutual understanding between persons of different perspectives, and opinions.

The Bandit Problem: Are You Winning Enough?

With the use of A/B Testing, two major challenges are faced by the online subscription industry today, i.e., to maximize conversions of trial users while testing product variants and to address ever changing user preferences. This project involves performance evaluation of Multi-armed Bandit Algorithms, as alternatives to conventional A/B testing, that balances exploration and exploitation during the learning process to quickly identify the overall winning feature variant. Contextual Bandit Algorithms are also explored to predict what works best for a given user based on their attributes. This approach helps optimize the conversion rate and support personalization at the user level.

The Smart Guide

Consumers are buying IoT devices without understanding their security and privacy implications. Traditional mainstream reviews only showcase device features. Consumers remain uninformed about security, but The Smart Guide hopes to change this. Our solution is a buyer’s guide for smart home devices. It helps consumers make an informed decision based on feature, price, and most importantly security. By factoring in cybersecurity in the rating of a product we hope to drive consumers to purchase safer smart home devices in a smarter way.

Tracking Global Corneal Transplant Performance

SightLife’s mission is to end corneal blindness by 2040. Millions of people around the world currently struggle to live with corneal blindness. It's a treatable condition with a clear solution: a corneal transplant. Our project is to design a system to simplify data entry and analysis for SightLife and their 32 partners so SightLife can better assist the partners in making decisions regarding cornea donations and cornea transplants. We are harnessing the power of a cloud based solution, using Azure to host our database, and PowerBI to ultimately improve data entry efficiency and allowing users to have flexible data analytics.

Understanding the Teacher Identity Development of an Undergraduate Data Science Practitioner Turned Educator

How to teach data science effectively is still largely unknown, as is the experience of educators in this field. Many data science instructors transition from practitioner roles, and understanding their experience during this transition may inform us of ways to support these individuals. To study this, I transitioned from data science practitioner to educator and tracked the experience with journal entries. The study found that I encountered many challenges, mainly surrounding issues of teaching in a higher education context, caring about others opinions, and difficulty distinguishing identity between peer, TA, or instructor.

iSchool Capstone

2019