General Insurance Pricing Seminar 2016 (7th June 2016)
I will be speaking at the General Insurance Pricing Seminar about Penalised Regression. To download the slides sign up here.
Penalised Regression refers to the family of methods for building Generalised Linear Models using Regularisation. Methods include Lasso Regression, Ridge Regression and Elastic Net. These techniques are commonly used in the machine learning community with applications such as Spam Filters and Genomics, but are not widely used in the insurance industry.
Penalised Regression is one of four machine learning algorithms that I have chained together to form a machine learning pipeline, that promises to beat any existing Pricing or Marketing Team’s predictive models – both burning cost and propensity. For more information on how we are transforming the way insurance companies build predictive models contact us. For an introduction to Penalised Regression click here and read the 2015 Actuary Magazine article “Judgement Day for Pricing”.
GIRO Conference 2016 (23rd September 2016)
I will be speaking at this years GIRO conference about Topic Modelling. Topic Modelling refers to the family of machine learning techniques designed to derive insight from unstructured text data. Using topic modelling we can
- Identify the common topics/themes in unstructured text data
- Visualise these topics using an interactive web based visualisation.
- Score new documents and assign them into one or more topics based on their topic probabilities.
- Feed the topic probabilities into a downstream predictive model
This methodology has clear applications for claims, fraud and customer service. To read how topic modelling helped a leading casino help understand their customer journey click here.