Predictive modelling in insurance is performed for many years by actuaries with the help of statistical models (e.g. Generalized Linear Models - GLM). The advantage of statistical models is that the final result is usually easily explainable (e.g. multiplicative form of GLM) by quants but also by other stakeholders (non-quants) involved in the decision-making process.

Machine learning techniques are now more and more popular in the insurance industry and have a lot of applications (pricing, reserving, claims management, underwriting,). Whereas the advanced techniques (e.g. random forest or neural networks) usually have a better predictive power than statistical models, their main drawback is that they are black-box and their results are difficult to understand/interpret which doesn’t always provide sufficient comfort to take business decisions. Hopefully several techniques have been developed the past few years in order to better understand the results of machine learning techniques.

In this presentation, we introduce (with a focus on their practical use and not on mathematical details) the concepts of features importance, partial dependence plots (PDP) or variants (M-plots or ALE), individual conditional expectation (ICE), Shapley value, H-Statistics for interactions, and explain how they can be used to boost insights from data in insurance applications (thanks to adequate features selection, features engineering and results interpretation).

These interpretability tools make the use of machine learning techniques much more relevant in insurance as it allows to improve the predictive power while understanding the drivers of the results; which is fundamental to take relevant business decisions.

We believe it is of uttermost importance that actuaries, who might use machine learning techniques in their daily jobs, also master these interpretability tools to improve the impact of insights they are discovering in data.

Find the Q&A here: Q&A on 'Artificial Intelligence and Machine Learning'