Description: The cloud inspires awe and fear in equal parts these days. There is a need to learn more, but there’s much information to learn and it may be difficult to know what to learn next. In this session, we’ll carve off a small corner of the cloud and make it your own. Setting up a Data Science Virtual Machine on Digital Ocean will be demonstrated. It’s the fastest way to get a data science development environment setup for your own use. Many of the topics and ideas covered will apply to almost all other cloud offerings. What use cases Digital Ocean is good for, and what use cases you should look elsewhere for, will also be discussed.

Description: Machine learning models are not black-boxes; these models can easily be interpreted using tools like SHAP, LIME and Anchor. These tools can facilitate the understanding of the model and the business problem. We can used these tools to build better models and communicate how predictions are generated. This article demonstrates their use in a mortality model.

Description: The session will feature case study discussion of how insurers can improve their model architecture and data visualization to prepare for the increasing demands of the future. Actuarial data and analysis has long been limited by the complexity and diversity of historical systems. External pressure on insurers is higher than ever before, as new requirements take effect under PBR, GAAP Targeted Improvements, and IFRS17.

Description: Ever wanted to do a Kaggle data science competition, but didn't know where to start? This presentation meets the needs of actuaries interested in predictive analytics and competitive data science. The presentation will describe the nature of competitive data science, and provide insights into lessons learned by competing in the Airbus Ship Detection Challenge at Kaggle, as well as other Kaggle competitions. The presenters will provide an introduction to neural networks and describe the nature of the Airbus Ship Detection Challenge and other Kaggle competitions, plus a solution approaches that can be used to obtain top scores using Convolutional Neural networks (CNNs), random forests and boosted trees.