Print this page
20 Oct 2015

Predictive analytics using social networks to anticipate the stock market

Social networks can help to predict the value oscillations of the shares, specially if the messages are negative.

It is not like a crystal ball of 140 characters, but it sets the tone, indicates trends and allows to predict with a surprising accuracy the future oscillations of a stock-market value. The social network Twitter has allowed to approach in a until now inimaginable way to the dream of instant information, and a large number of experts and analysts believe that behind it the secret of wining on the stock exchange is hidden.

A study published in 2010 after the analysis of 9.7 million messages proved that the possibility to find out in a 87,6% the following evolution of the Dow Jones Industrial Average based on the messages send in twitter by the investors exists.Another research, based on the spanish Ibex 35, shows that the published messages of this social network influence the following evolution of the values mentioned in at least the 69% of the cases.

"In some circumtances and for determinated values, social networks provide very useful information that can inform about the deviation of this value on the market average and, thus anticipate the evolution of the value", Raul Gomez Martinez,CEO of Apara, explains. Furthermore, Mr. Gómez is professor of financial economy at the University Rey Juan Carlos I and spanish research responsible, and according to its conclusions, the influence of each message with positive or negative content come to exceed the 50 basic point of its price in certain occasions.

Futbol and summer influence

The impact of the exogenous factors on the investors' mood and its following effects on markets has been extensively analysed in several studies. Factors as the climate, the period or the sports results cause oscillations. Researcehrs think that the key is to capture and interpret rapidly the mood of investors and thus anticipate their following movement on the market.

"The problem has always been on how to measure the investors' feelings. For intance, surveys are expensive and slow, and when obtaining the results, the market moment has already passed. However, nowadays, social networks and big data allow to follow it instantly."Gómez said, he is about to publish another research that shows that a published "tuit" about a value can cause a deviation up to a basic point of its price.

The research team of the Rey Juan Carlos University analyzed the 35 values of the Ibex during a 6 months period capturing the opinions that twitter users send out about the values that form the spanish selection with the objectivo of capture the market expectations. Daily,the tuits that mention the Ibex35 and its values are collected and narrow down to remove retuits and the messages with non valid words. Then they are classified according to if they are positive, negative or neutral and they are given a score according to criterion as the relevance of the user or the quantity of retuits, in a process that is finally checkes by an administrator.

According to researchers, when the correlation between the tuits and the value of the shares is when the messages analyzed are negative, precisely because in these cases more information is generated. "The negative feeling is more representative. It is a psychological thing. If you are fine at work you don't tell to anyone. Conversely, if you are annoyed you tell to everybody" explains Gómez.

See the complete article on