yangvincent / predict-elections

Predict the California Primary 2016

Predict Elections

About

This website uses polling data provided by the WSJ coupled with census data to more accurately predict the 2016 California Primary between Hillary Clinton and Bernie Sanders.

The server uses node.js to route different queries to/from a sqlite database and various API calls. The front end is completed with vanilla JavaScript, HTML/CSS, and D3.js.

Instructions

You can visit this site through the Heroku link or by running the server locally. Due to Heroku's new free tier, it may take a few moments to fully wake up the dyno.

In order to run site, run npm install on the needed modules, then node index.js to start up the server. The databases are stored in polls.db and census.db. These can be set up server side with node createDB.js, though with the .db files they are technically already set up.

1. git clone https://github.com/YangVincent/predict-elections.git
2. cd predict-elections
3. npm install node-static
4. npm install sqlite3
5. npm install request
6. node createDB.js
7. node index.js
8. Point your browser to localhost:5000.

Model

I took gender, age, and race into account and tried to predict what the outcome of California's voting would be. I weighed gender to only have 10% of influence on the final result, age to have 40% since most people from different age groups have drastically different opinions on controversial issues, and finally 50% for race because I feel that people are extremely affected by race, which correlates closely with peoples' upbringings.

Further Explanation

Since this assignment focused on integrating results from the Huffington Post API, I first had the browser side execute an XMLHttpRequest query to the server. The server then sent a GET request to the Huffington Post API, and waited for a result. Finally, when the Huffington Post results returned, it sent the results back to the client. After the client finished processing the Huffington Post results, it executed another query to figure out results for the combined model to display as well. This query also used fs on the serverside to save results for voting percentage predictions.

Warning

Because this data pulls the 3 most recent polls, some of the top results may be out of date and/or inaccurate. Specific predictions from the time can be found in Vincent.Yang.prediction.txt.