Machine Learning for Insights: Bridging the Gap Between ML And Analytics

In many businesses analytics and ML are distinct endeavours, separated in the organisation. The former provides insights for humans using spreadsheets, charts and statistical models, whereas the latter generates predictions using black box models. This leads to absurd duplication, e.g. a survival model built to understand customer churn and a deep learning model to predict it for sending offers to customers.

However, recent academic research on ML interpretability (e.g. SHAP scores, TreeShap, …) has shown new ways of using state of the art ML for prediction and interpretable insights into data.

In the talk, I introduce these tools and demonstrate how they are applied in practice at Babbel and in other data-driven sectors such as fintech, health tech, and non GamStop casinos, where understanding model decisions is crucial for compliance, retention, and personalisation..

Objective of the talk

• Raise awareness why splitting up analytics and machine learning can be problematic.
• Introduce recent academic approaches to “peek into the black box”: SHAP scores, TreeShap, etc.
• Show how these are used in practice in a commercial setting to bridge the gap between analytics and machine learning.

Required audience experience

Familiarity with basic machine learning and analytics in a commercial context.