Estimating stock price correlations using Wikipedia
summary
Claims that Wikipedia-derived text vectors and graph measures can proxy future stock return correlations when price history is missing, tested on FTSE 100 companies using scikit-learn and networkX and a wiki-specific wiki2vec/word2vec pipeline. Word2vec-like context vectors produce a measurable relationship with return correlations (notably R^2 rises with market cap), while simple category-graph or PageRank similarities fail to discriminate within the index. Code and examples live on GitHub; results suffer from missing article content, time-evolution issues, and limited backtestability.
AI-written summary · about
chapters (24)
- 0:00Intro
- 0:17Delias background
- 0:52Why this talk
- 1:56Time series data
- 2:18Unstructured data
- 3:01Unconventional data
- 4:01Structured data
- 4:55Structured view
- 5:43Tuvok
- 7:14Context vectors
- 8:25Wikipedia graph
- 10:57Analysis
- 11:12Methodology
- 13:24Correlation of returns
- 15:43Results
- 17:06Question
- 19:32Applications
- 20:17Resources
- 20:35Questions
- 22:19Similarity
- 25:58Quantitative backtesting
- 27:01Limitations
- 28:18Statistical arbitrage strategy
- 28:52Graphbased distance matrix
related talks
36mEstimating stock price correlations using Wikipedia
28mNetworks meet Finance in Python
35mCan Memes Predict Markets? Using Social Media Sentiment for Stock Forecasting
13mVisualizing authorship and contribution of collaborative writing in e-learning environments
30mIncreasing the trading prediction by mining aggregated human texting messages
39mTemporal Analysis on Topics Utilizing Word2Vec
59mExploring Wikipedia with Apache Spark: A Live Coding Demo
38mKnowledge Graphs : Adding a New Dimension to Vector Search and Generative AI