Two key technologies driving Machine Learning in Financial Services

Predictive powers

Many people wish they could predict what will happen next in the world. Many predictions are assigned to the waste bin of time very quickly. With hindsight, unforeseen factors come into play that changed their 'models'. It is because there were so many factors involved to predict. The ability of models to analyse and interpret means technology was not able to process, analyse and predict with a high degree of success.

However, take a narrow domain or area in financial services and predictive analytics can be very effective with current technology. Take Black box technology in cars polling back large data sets to insurance companies to analyse. Mix with data on weather, road conditions and driver claims history and demographics. Predictive models can quickly be created to determine key connections between driving styles and claims history. Predictive models degrade over time so continuing to refine with continual feedbacks of driving data, claims, and using machine learning techniques is critical to make and improve accurate predictions

Analysing Text

Another challenge in the era of big data is reading and summarizing huge amounts of unstructured data in emails, blogs and articles. Take an insurer of an oil tankers moving past the Somalia coast. Does the underwriter have time to read every report, local news article or email on the topic? Text analytics allows you to extract semantic information and sentiment from these articles in a structured form that can be searched and understood. A risk or sentiment score can be assigned that brings to the surface critical information around insured assets and allows an employee to focus on critical risks not information noise. Put Text analytics together with natural language chat bots and you have the ability to make simple inferences and answer questions about the unstructured data. As time goes on the models improve and are refined to better reach more accurate answers. Analyst such as Mckinsey predict such use of intelligent chat bots could mean al large percentage of call centres roles obsolete within a decade.

However, these standalone technologies of predictive and text analytics can create more value when inserted in business process applications that create transactions data to create Intelligent applications. SAP provides new wizards to quickly create predictive models and automation tools to train these models on data over time. When inserted into SAP business applications it can quickly create additional business value. For instance, when embedded in claims management systems premiums can be quickly altered to reflect the true risks and aligned revenues with exposures.

So expect more rollout of predictive analytics with machine learning capabilities embedded into applications in customer, risk , finance or claims over coming years. In addition, the ability to embed into existing applications to add more value. Natural language processing and text analytics of unstructured data is also going to drive a large improvement of productivity of many customer service and knowledge management roles over the next decade. Many of these new advances will free up time to spend on more value adding and reviewing roles but I do expect some role casualties over time. I'll be discussing the key benefits, challenges and concerns around the spectrum of machine learning at the upcoming webinar, "Making Sense of Machine Learning In Financial Services" on 28 March 2017. If you'd like to get a better understanding into what these technologies mean for you, you can register here :https://www.finextra.com/featurearticle/2185/webinar-making-sense-of-machine-learning-in-financial-services