UCSD MAS-DSE 260

Mapping the U.S. Non-Profit Space : Classy Data Analysis Project

Advisors: Ben Cipollini and Ilkay Altintas

CONFIGURATION

make build
make start
make stop

MODULES

In broswer window go to: localhost:8888

APP

In broswer window go to: localhost:80 or localhost

INTRO

The social sector is typically viewed in terms of nonprofit organizations and the cause categories they belong to. It’s clear, however, that while younger generations are active in social causes, they think more in terms of current events and social causes organizations - so much so, that new donor churn now peaks at over 80%.

It is not clear, however, how to lay out a common “social space”, where the organizations that drive social change and potential donors could connect, find organization and cause recommendations, and where discovery of new causes - and news events within causes - could be facilitated.

In this project, we build this social space from the ground up. We use a combination of government IRS form 990 (returns for nonprofits) data along with external textual information (i.e. social media) to create a robust semantic space.

TEAM

Budget Manager: Jeet Nagda Management of AWS funds and cloud-related activity Project Manager: Howard Tai High-level project planning and assignment of responsibilities Project Coordinator: Erin Hansen Stakeholder correspondence and record-keeping during meetings Report Manager: Carlos Pimentel Tracking and coordinating report deliverables Record Keeper: Juan Reyes Management of Github repo and Docker environment

ARCHITECTURE

Our data pipeline was divided into different modules to accomplish discrete tasks. In our final product, we used Python and Docker throughout as our common language and platform. The purpose of our first module (Form Data Processor) was to gather and fetch information from the AWS repository using the IRSX tool for each of the organizations in a giant manifest. We had an additional experimental module (Website Data Processor) which additionally scraped the HTML text of the organization’s website listed on the form 990. The XML payload was parsed and stored as a document in a MongoDB instance. We then had a clustering module (Cluster Processor) which read data samples from the MongoDB instance to create three labels or cluster IDs for each organization: one for each axis of comparison. These labels were loaded back into the MongoDb document for persistence. Finally our last module (App Web Server) reads from the database to create visualizations for high level queries, or a given input organization

DOI

Identifier: doi:10.6075/J079431XIdentifier https://doi.org/10.6075/J079431X

Creators:

• Tai, Howard;
• Hansen, Erin;
• Nagda, Jeet;
• Pimentel, Carlos;
• Reyes, Juan

Title:	Classy Data Analysis: Mapping the U.S. Non-Profit Space
Publisher:	UC San Diego Library Digital Collections
Publication year:	2019
Resource type:	Dataset/Dataset