Insights into Social Media Data using Entropy Theory
summary
Applies entropy measurements toTwitter streams and claims entropy spikes precede engagement spikes while entropy falls predict imminent collapse, with the most consistent result being predicts trend dying out 3–4 days early. Implements a Python pipeline that fetches and cleans JSON into CSVs, uses a library called thwack, and tests on 80 million tweets across seven datasets. Produces visualisations and regression tests, hits memory limits, and finds early‑detection works on historical data but not robust real‑time prediction yet.
AI-written summary · about
chapters (8)
- 0:00Introduction
- 0:25Welcome
- 2:02Why this project
- 7:13Analogy Applications
- 14:43Objectives
- 17:18Collecting data
- 19:08Python Pipeline
- 21:57Preliminary Findings
related talks
34mUsing Twitter's Breakout Detection Package
01h 05mSocial Text, Sentiment, and Tone Analysis
11mThe Value of Social (for) TV
27mExploring Science on Twitter with IPython Notebook and Python Pandas
09mSocial Data and Television
56mSocial Listening, Data Extraction & Discovery
41mDemystifying Social Network Analysis (SNA)
38mAnalysing Russian Troll Tweets data with Python