We have shown in previous posts, how, the new data sets that are available can be a key source of trading opportunities. In addition getting access to this information before it becomes ‘news‘ on the mainstream media is highly valuable.
However with there now being over 277,000 tweets, 571 websites created, 347 new blog posts and 72 hours of new video being added to YouTube every minute of every day, the volume of data to read and filter is growing exponentially.
So how do you find and alert people to the most relevant information, amidst all of this ‘noise’?
Utilising proprietary systems, Heckyl has built its own semantic platform, to structure, read and filter the vast data sets available. Taking the principles that have been applied to event driven trading, our system treats every blogger, Twitter user etc. as a data point to be analysed. By back-testing the key events of all of the instruments that we cover – more than 35,000 at last count – over the last five years, we check the related content from these millions of sources for causation and correlations.
This enables us to create a relevancy score for each source, be they journalists, analysts, twitter users or bloggers, as well as understanding the types of information these sources have a history in predicting. This automated process utilising machine learning algorithms means that we are able to filter and rate hundreds of millions of daily data points, in real time, to alert our users to sensitive financial information. We separate the good data from the junk by giving scores to the users as well as the content, where only the scores that meet our threshold is selected by our algorithm.
One recent example can be seen in an alert we generated from the blogger Brad Reese on 7th August this year. Brad is given a high rating by our system in relation to Cisco news and events, with a score of over 80, calculated on the basis of all his previous posts.
On 7th August he posted a comment about the layoff of 6,000 staff at Cisco’s Data Centre division, by October 2014. The news was officially announced by Cisco on 12th August, which is when the overall sentiment dropped and there was a 5% fall in the share price.
Finding that ‘needle’ amongst the collective noise can prove to be very valuable.