Discover how to use advanced techniques in Python, including time series analysis and the Lee-Carter model, data cleaning, and visualisations to forecast mortality rates in this end to end walkthrough of mortality modelling.
.jpeg)
If you’re looking to learn how to utilise advanced applications of Python in an insurance context, this course is ideal for you. An individual subscription gives you 3 months’ online access to:
As Well As
Our Industry and Actuartech Resource Libraries which feature curated additional content to assist you on your data science journey.
You can also request to access to a coding project to practice the skills you learn in this course.
This case study aims to show how we can analyse and forecast mortality in old ages by illustrating how the Lee-Carter model and Cairns-Blake-Dowd model can be fitted on mortality data. As the packages for fitting Lee-Carter and Cairns-Blake-Dowd are fairly scarce, we will be building the models from the ground-up.
Using Python techniques discussed in our foundations in Python course, we are able to construct an algorithm and generalise it using classes as well as generate forecasts. The end result is a set of usable functions and classes (built from first principles) that the user can apply to other mortality data.
This course aims to:
We explore Long Short Term Memory (LSTM) neural network approaches to time series forecasting using different packages. The results from the LSTM are compared to the Lee-Carter model’s random walk forecasting technique.
If you're not already familiar with Python, we recommend that you start with our Foundations in Python course.
In Chapter 1 we define the problem we wish to solve and provide background into the models we will be fitting.
In Chapter 2 we outline the relevant packages we will be using and we explore the mortality dataset by validating that it is complete before continuing with visualisations.
Chapter 3 sees us outlining, from first principles, how we fit the Lee-Carter and Cairns-Blake-Dowd mortality models. We collect the algorithms into a single class that can be called (per model) which makes analysis easier than re-running multiple code blocks.
In Chapter 4 we use deterministic techniques, and we forecast both the Lee-Carter and Cairns-Blake-Dowd fitted models. We then explore simple stochastic techniques by adding a variance component.
Lastly, in Chapter 5 we conclude with an introduction to LSTM techniques to perform time series forecasting and compare these techniques to the Lee-Carter model.
We have tailored packages available to ensure that corporate teams have the option to attend structured live lessons by our tutors, and the option to request a practical assignment and bespoke feedback. Invoicing option available.
