Gilbert et al., ICWSM 2010

This a Paper discussed in Social Media Analysis 10-802 in Spring 2011.

Citation

Title : Widespread Worry and the Stock Market Authors : Eric Gilbert, Karrie Karahalios Conference : ICWSM 2010

Online version

Widespread Worry and the Stock Market
Website for the paper identify the latent class as well as dyadic events

Summary

The stock market usually reflects business fundamentals, such as corporate earnings. However, we also see many events that seem rooted in human emotion more than anything else, from “irrational exuberance” during booms to panicked sell-offs during busts. In this paper, Gilbert et al attempt to explain the downward movement of stock prices using sentiment analysis methods. He build an anxiety index for LiveJournal, an active blogging community. Then, they correlated their anxiety index to the S&P 500.

Brief description of the method

There are two main parts involved in predicting stock market movement base on anxiety

• Boost Decision Tree
• bagged complement Naive Bayes algorithm

After training on the data set, the author used a combination of these 2 classifiers to assess the anxiety level of the blogging community.

To check accuracy, the author used Granger-causal Analysis to test whether anxiety level gave good information to movements on stock market.

Experimental result

The resulted predictions are tested using Granger-causal analysis, which use econometric techniques to tell whether the Anxiety Index provides useful information for projecting future stock market prices not already contained in the market itself. Statistically, this paper showed that in general moods from an online community has novel predictive information about the stock market.

Dataset

LiveJournal post stream gathered using atom listener on: LiveJournal atom stream LiveJournal post stream gathered using atom listener

Related Papers

There has been a lot of work on using sentiment analysis to predict the stock market.

• The paper by Chua, Milosavljevic and Curran Uses bayes classifiers to classify the sentiment of a internet stock message board.