What They Measured (And How They Got the Data)

Intro

Section of this node - https://www.hubnx.com/nodes/200b2ec1-37ec-4be6-adc7-57838917d255/sections/c1677a39-c987-48cd-8d2f-c896a09eff0a

In this node we would discuss how researchers gathered the data, the strategy they used, and what exactly did they mesaure to determine the sucess of movies? We would make sure to highlight the important details, but will not complicate it!

Collection of Data

The researchers picked 24 movies released between November 2009 and February 2010 - including big names like Avatar, The Blind Side, and Twilight: New Moon.

They collected:

• 2.89 million tweets over 3 months
• Data using the Twitter Search API
• Keywords = movie titles
• Time range = 7 days before release → 14 days after

Then they measured two key features:

1. Tweet-Rate

How many tweets per hour each movie got, before and after release.

Example: “Twilight: New Moon” had over 1,300 tweets/hour before opening.
“Transylmania”? Just 2.75/hour.

This variable alone turned out to be incredibly predictive.

2. Sentiment

They built a sentiment classifier using Mechanical Turk.

• 3,000 tweets were labeled as Positive / Negative / Neutral
• This became training data for a Naive Bayes classifier
• It was then used to analyze sentiment trends for all 2.8M tweets

They also collected box office performance data and number of theaters released in, to add more context.