Needless to say, our “data geeks” have some serious brainpower. Hear from our data scientists, Ayodele Odubela and Daniel Laney, as they provide some valuable insight into what’s on an MVR.

Highlights from the webinar include:

  • Risk modeling is the process of using relevant historical data and domain expertise to understand the probability of future risk events.
  • SambaSafety focuses on entity resolution, which is the process of identifying and linking different manifestations of the same real-world object.
    • Since they are trying to predict risk, this means that behaviors collected from different sources are pulled together to understand and predict risk.
  • As far as the sample design goes, you can build predictive models.
    • Sample design is extremely important so it accurately reflects the real life decision making process.
    • What do they do the first time they look at an MVR?
      • They physically look at the document and make a judgement – all manually. We are doing the same thing but just with machine learning
    • With sample design, there is observation periods that represent aggregate data input into the statistical machine learning models.
      • The second component is the performance period, or what actually happens.
    • With an observation period, it only looks backward in time.
      • This is all historical data that has been deemed relevant, was collected and aggregated into features.
      • These features are then used to make risk profiles as well as predictions.

To learn more about topics including target selection and how that access to performance data allows us to develop target variables, supervised learning and modeling, watch the webinar on-demand.