WhaleStreet Provides Stock Exchange Indicators Based on Twitter Data

By June 08, 2012
Keywords : Smart city, Europe

This startup company offers a predictive trend indicator based on analysing ‘tweets’ to gain an understanding of how Twitter users feel about a given subject. It works by making use of historical correlation analysis.


A report from the BI Norwegian Business School recently claimed that fluctuations in the Stock Exchange share price of a given company could be correlated with the general feeling people have about that company, as conveyed by their Twitter posts. Nor is the BI study the only work being done in this area. Startups such as WhaleStreet have begun to launch services on similar themes. This young French company has just been chosen to take part in the third season of France-based incubator Le Camping which describes itself as a “Kickoff for Startups”. Its newly-developed platform, also entitled WhaleStreet, enables real-time gathering of tweets and then derives indicators by applying statistical, predictive algorithms. Alexandre Hajjar, CEO and co-founder of WhaleStreet, explains that these indicators can “predict the movement of a share price over the following one or two days, based on the general sentiment, the feelings, of Twitter users and then verify it.” The idea is to maximise the accuracy of the analysis by using a large number of messages – first and foremost in English, as this is the language most often used on social networks, plus French.

Feelings matter

How the system works in practice is that a customer who signs up to WhaleStreet’s services will receive information based on a correlation between the general feeling, as expressed over Twitter, on a given subject – a company, a range of products, an event, etc – and a share price on the Stock Exchange. The tweets are filtered using keywords that align with the chosen subject and the "system enables the user to read the sentences automatically using semantic analysis, and so to understand the feelings, the positive or negative sentiment being expressed," explains Hajjar. And this appears to be information worth having given that, over the five months they have been running the system, the founders have seen its predictive accuracy improve by 18%. Of course, in order to make this work, "we have to apply to this mass of data a learning algorithm able to make the connection with past correlations," adds the WhaleStreet CEO.

Learning from the past

The algorithm analyses a past correlation between a given sentiment as expressed on Twitter and a change in the share price quoted on the Stock Exchange. From this correlation it derives a real time indicator which could then be transmitted by a newsletter alert service or via a message on the screen of a computer, smartphone, etc. WhaleStreet thus provides a basis for deciding whether or not it’s a good time to buy or sell a portfolio of shares. However, comments Alexandre Hajjar, "recommendations coming from the algorithm do retain an element of uncertainty; they should be seen as additional information to supplement other data sources." However, in the longer term, he claims, WhaleStreet "could be used to make predictions about things that are likely to happen”.

Legal mentions © L’Atelier BNP Paribas