Machine learning is a branch of artificial intelligence whereby computers use algorithms to analyse data to identify patterns and build predictive models. With a minimum of human intervention, the models are automatically adapted to produce increasingly reliable and insightful results as they are exposed to new data. This type of computer-driven intelligence is transforming how global procurement departments uncover and proactively address risk—from vetting potential vendors to on-going supplier monitoring for reputational, financial or regulatory red flags.
How Machine Learning Works
The term ‘machine learning’ was coined in 1959 by Arthur Samuel, an American pioneer in computer gaming and artificial intelligence, while he was working for IBM. In classic machine learning, methods are applied to statistical problems with well-defined and structured datasets. This can be taken further to the point of ‘deep learning’, which mimics the layered learning process in the neurons of the human brain. Algorithms are stacked, with one algorithm lifting a certain feature from the data, which is fed to the following algorithm, which lifts another feature, and so on. Rather than these layers of features being designed by human engineers, they are learned from the data by means of a general-purpose computing procedure.
Due to its reliance on large datasets and heavy computing power, machine learning is closely associated with the more recent big data revolution, says Bart van Liebergen, associate policy advisor at the Institute of International Finance, in his article Machine Learning: A revolution in risk management and compliance? “While elements of machine learning go back to the early 20th century,” says van Liebergen, “widespread use picked up as computing innovations and greater availability of high frequency data allowed it to model complex, non-linear relationships, while making machine learning much easier to be applied.”
Even if stacked algorithms, structured datasets and predictive analytics are challenging to understand for the average person, the value is easy to see. Self-driving cars; personalised online product recommendation based on previous buying histories; analysis of trends and data from wearable monitoring devices are all applications that improve our lives.
Likewise, business applications are enhancing organisations’ ability to better mitigate risk—from predicting refinery sensor failure and streamlining resource distribution in the oil and gas industry to identifying important insights in its masses of transactional and regulatory data, and the prevention of fraud and in the financial services industry. Machine learning can also help with identifying risks associated with bribery and corruption or forced labour in supply chains.
“One area in which machine learning has been applied for more than a decade, and with significant success, is the detection of credit card fraud,” says van Liebergen. “The historical transaction datasets showcase a wide variety of pre-determined features of fraud, which distinguish normal card usage from fraudulent usage.”
Machine learning provides valuable insights because it can identify complex patterns in data faster than human analysis. This empowers organisations to use both internal and third-party data to gain a more comprehensive understanding of risk among customers, suppliers and other third parties. With such insights, organisations can respond proactively when a reputational, regulatory, financial or strategic threat appears on the horizon.
Why is Third-Party Data Critical to Machine Learning?
Organisations have large volumes of data already at their disposal from internal sources across the enterprise—from data captured in financial systems and CRMs to data captured by smart machines in production lines. But internal data alone will not power truly insight-delivering machine learning applications.
What type of third-party data helps develop a holistic picture of risk?
- Negative news from print and web-based sources to stay on top of emerging threats
- Country and industry reports to predict strategic risks and opportunities
- Business data including financial details to identify bankruptcy risk
- Corporate hierarchies to help uncover beneficial ownership
- Sanction, watchlists and politically exposed persons (PEP) lists to mitigate anti-bribery and corruption or anti-money laundering and counter-terrorism financing risks
- Legal information to help understand litigation histories of potential customers or vendors
Machine learning is unquestionably a powerful vehicle for advances in the field of predictive analytics. But for corporate due diligence and risk management—including the many and ever-increasing regulatory requirements impacting global business—this “machine” runs most effectively when given the right type of fuel.