he transition is underway in everywhere of the world with many companies including insurers starting up specialist “data science” teams. Machine learning – a trademark skill of a data scientist will soon become the preferred method of building predictive models and creating data analytics solution. Point and click Generalised Linear Modelling (GLM) – the current insurance industry standard for risk modelling – had become the “old way of doing things”.
The Asia predictive analytics challenge created by Big data working party from Singapore Actuarial Society in 2016 aims to create more awareness of machine learning, data science and technology in insurance industry. In the challenge, we were the winner of the competition among actuaries.
In this data case study, we would like to share with audiences how we approach the challenge, step-by-step workflow and some of the secrets we use to win. We would also like to show how standard statistical analysis and GLM models are not as competitive as machine learning in comparisons.
There’s also a popular claim about the interpretability of machine learning models: Simple statistical models like logistic regression yield interpretable models. Neural networks, on the other hand, are black boxes. This brings us to a question of trust: do I trust that a certain prediction from machine learning model is correct? Or do I even trust that the model is making reasonable predictions in general? We will show some of the ways we use to gain trust from the models and how do we make an informed decision after the black-box machine learning model is rolled out in the real world.
©2017, O'Reilly Media, Inc. • (800) 889-8969 or (707) 827-7019 • Monday-Friday 7:30am-5pm PT • All trademarks and registered trademarks appearing on oreilly.com are the property of their respective owners. • email@example.com