Machine learning has been in the news recently thanks to Google’s AlphaGo programme’s success at playing Go – the 3,000 year old Chinese board game. The system learnt the basic rules of the game and then effectively played millions of games and learnt from its mistakes each time.
But all kinds of businesses are now using these sorts of systems in everything from web analytics to ensuring compliance with financial and fraud regulations.
Essentially machine learning is the flip side of big data. There’s not a lot of point in collecting tons of business information unless you can turn that into useful, actionable knowledge.
To do this in a timely manner you need some help – that’s where machine learning and artificial intelligence comes into play – finding the needle in the haystack of data.
For instance Hewlett Packard Enterprise offers software to help financial services companies spot fraud and dodgy transactions. This market needs help because it has to deal with ever more complex regulations which have been changing almost weekly since the crash of 2008.
Simply keeping up with the increasing rules is very difficult for IT and compliance departments.
By automating at least some of these functions the whole process can be made far more manageable. The latest systems can spot suspicious trading linked to any kind of market announcement or even activity on email, LinkedIn or other social media. Any trading that might cause concern can be highlighted for further investigation.
There’s an interview from HPE Discover Las Vegas discussing artificial intelligence and financial services here: https://youtu.be/WA_Oz8NFsAI
Financial institutions, with their big IT investments and infrastructure have been among the first sophisticated users of machine learning. Banks and other organisations are not only using machine learning for compliance but also for more profitable trading strategies.
There are systems which analyse vast quantities of historical data in order to predict future market changes. Some look for discrepancies in prices – if one company in a sector is trading much lower, or higher, than comparable firms for instance.
But machine learning systems can also support human decisions by improving the way trading is done.
So after a decision is made to buy a certain stock then the system advises on the best way to make that trade. Once trading is finished the system can analyse results to make even better decisions next time.
Companies also use machine learning systems to provide strategies to hedge risk – to find investment strategies to mitigate possible downsides of other investments for instance. Or they can follow competitor strategies to either mimic their strategy or to bet on the impact their trading will have on the wider market.
But machine learning is also playing an ever deeper role in other industries.
One new area is in studying what has previously been seen as ‘unstructured’ data which has traditionally been viewed as too chaotic to be useful to business.
Social media creates massive amounts of this unstructured data which business has found impossible to analyse.
Systems can dredge the thousands of social media posts, tweets, Instagram images and Facebook updates to try and gauge the impact of advertising or marketing campaigns for instance.
They can crunch the numbers created by web analytics programmes to help improve how home pages are designed or specific campaigns are run.
This promises a more profound impact for the business world because most companies create far more unstructured than structured data – and the vast majority have not seen this data as a resource which they can exploit.
But even more profound is the next step for machine learning. Businesses are beginning to get help from computers not just to judge the effectiveness of past actions but to suggest changes to future strategy.
Machine learning has been used by many businesses to look at how they performed in the past, but the next chapter will see some businesses using machines to suggest what they should do in the future.
