In this newsletter...
VIRTUAL EVENT 22 June: Practical Steps to Enhanced Pricing in Non Life Insurance: Request to join here
LIVE WEBINAR 25 June: Introduction to Machine Learning to understand Motor Insurance Claims - register here
LIVE WEBINAR 30 June: Investigating Applications of Data Science in UK and non-UK Actuarial Teams
Insights Article: Building a Simple COVID-19 Dashboard using RShiny
JOB OPPORTUNITIES - see below
OUR DATA SCIENCE COURSES - register here or contact: firstname.lastname@example.org for a DEMO and corporate rates
IFRS 17 TRAINING: HANDS-ON Introduction Course Subscribe here
For more information about any of the above please contact us
Introduction to Machine Learning to understand Motor Insurance Claims
Friday 25 June 2021 08:30 AM BST
On the 25th June 08:30 AM BST, we are pleased to present Introduction to Machine Learning to understand Motor Insurance Claims.
We will show you how to utilise data science techniques to analyse motor insurance claims.
By fitting advanced data science models we aim to predict the frequency of these claims to assess risk.
This case study is designed to give a broad overview of basic machine learning techniques that can be used to model data in the context of a motor insurance company.
These techniques will be applied to a particular dataset (see below) in conjunction with more traditional actuarial techniques, including a comparison of performance in both instances with the results produced explained. In addition, we will touch on how certain models could potentially be fine-tuned to yield better and more explainable models, in the context of this dataset.
We will be using a dataset of motor contracts from a French insurer.
The file is named ‘freMTPL2freq’ and can be found on kaggle.com here.
During this introductory webinar, together with Reacfin we will show you:
- Overview of how we utilise data science techniques to analyse motor claims using the dataset above.
- How to fit a machine learning model that aim to predict the frequency of these claims to assess risk.
- We demonstrate how these techniques could be applied to improve reserving and pricing processes within motor insurance.
This webinar will be presented by:
- Xavier Maréchal, CEO Reacfin
- Patrick Moehrke, Junior Actuarial Consultant at Dupro Advisory
- Julien Crespy, Actuary & Data Scientist
Investigating Applications of Data Science in UK and non-UK Actuarial Teams
Wednesday 30 June 2021 08:30 AM BST
On the 30th June 08:30 AM BST, we are pleased to present Investigating Applications of Data Science in UK and non-UK Actuarial Teams.
Significant changes in technology, regulation, markets, customer behaviour, the environment and other global trends are influencing the actuarial department. The increasing availability of big data, The availability of technical data science skills, and the application thereof; are changing how insights are being derived and continuing to shape the operating model of the actuarial department.
We performed benchmarking exercises which involved structured interviews with senior first line actuarial department representatives from different UK, South Africa, Belgium, Luxembourg & Switzerland life and non-life insurance organisations to investigate how the insurance industry is utilising data science, with a focus on application and use cases within an actuarial context. We investigated the strategy and the operating model within which data science is used including the types of tools and techniques being used.
Within our benchmarking exercise we also included themes around the types of data; the technical nature of machine learning techniques and software being used; and wider considerations including risks, risk management, governance, and ethics related to data science.
We investigated trends impacting the skill set required by those working within Data Science and the barriers to adopting data science.
We interviewed representatives from first line actuarial departments mainly with Heads of Actuarial Reporting and Pricing Departments, Heads of Actuarial Systems and Heads of Actuarial Transformation and Strategy including direct insurance organisations and group entities.
Our talk will summarise the findings from this Actuarial Data Science benchmarking exercise.
This webinar will be hosted by Actuartech, Reacfin & Synpulse.
Valerie du Preez FIA, Managing Director, Actuartech & Dupro Ltd
Xavier Marechal, IA|BE qualified actuary, CEO Reacfin SA
Anja Friedrich, Actuary SAV, Manager at Synpulse Management Consulting.
LIVE VIRTUAL EVENT
hosted by Quantee & Actuartech
Practical Steps to Enhanced Insurance Pricing | Tuesday 22 June 2021 08:30 AM BST
1. Introductions and industry insights
2. Three steps to enhanced insurance pricing
3. Using technology in your pricing journey - demo
Dawid Kopczyk, FIA Quantee, CEO
Valerie du Preez, FIA Actuartech MD
If you are interested in joining us for this Llive Virtual Event, please request access here
Building a Simple COVID-19 Dashboard using RShiny
Introduction to Dashboarding in R with the use of COVID-19 Data
With the rise of the COVID-19 pandemic, there presents an on going need to monitor data regularly in order for businesses to interpret and respond to the impact of the pandemic. This presentation is concerned with actuarial teams; and how they may use lightweight dashboarding techniques aspart of an internal reporting strategy; allowing them to understand or forecast trends; understand the impact on their business and share information required to these impacts.
Our aim is for users to have a set of relevant pages displaying key COVID-19 statistics interactively, like the ability to change the country, date, and statistic they wish to view. To make it more readable, we want to display these in separate tabs. One page displaying COVID-19 statistics at a glance, labeled as "Overview", another displaying an interactive table the user can search through ("Table"), and finally an interactive plotting environment where the user can compare countries across various features (e.g. daily deaths, cumulative tests conducted). We also want to give users the ability to add moving averages. This smooths out the visualised data while preserving trends.
R Shiny Dashboarding
Implemented correctly, a dashboard should be intuitive to the user to operate and produce clear visuals and insight from data. The user typically will not have to add code nor require knowledge on how to code, as the code operates in the background. This offers users more flexibility than static visualisations but without requiring the users to know how to produce visualisations from scratch.
To implement a Shiny dashboard in R requires a shift from traditional programming structures, known as reactive programming. This essentially means code can be run and objects may change in real-time while the program is running. To contrast this to usual R programming, when we run the code to train a model and produce a summary of the output, we cannot alter that model while it is running and see results in real time. It requires that we run the piece of code again before changes can be seen. With Shiny's reactive programming, we only need to run the R script once to load the dashboard, then by interacting with the dashboard, we "re-run" part of the code again with new changes made. As you will see, we are able to switch from country to country and alter the dates and the output (for example: case numbers, deaths, and vaccines) update automatically
Figure 1: The dashboard provides key insights at a glance through a web browser
COVID-19 Dataset Used
We utilise the most recent version of Our World in Data's COVID-19 dataset, found on their GitHub repository https://github.com/owid/covid-19-data/tree/master/public/data/.
The dataset contains the following information relevant to our dashboard:
1. Reported Cases and Deaths: Daily and cumulative values for each country from the first reported case worldwideto present day. This is provided from the COVID-19 Data Repository by the Center for Systems Science andEngineering (CSSE) at Johns Hopkins University.
2. Reported Tests: Daily and cumulative amounts compiled by Our World in Data.
3. Vaccinations: Daily and cumulative amounts compiled by Our World in Data using published reports. Before being used in the dashboard, the dataset is checked for quality and structure by checking for missing values,obvious outliers, and visualisation spot-checks.
In addition to the R programming language, we also require external packages that allow us to build the dashboard, namely shiny, shinydashboard, DT, lubridate, tidyverse, and zoo.
Note that this is performed using R 4.0.3 for Windows 10 x64.
Loading and Pre-Processing the Data
The first step is to load the required packages and data onto the local system. Any missing values are set to zero and we ensure the dates present are Date objects. The data is then stored as a reactive object.
The output from user selections will be captured in the server function and updated in the ui object.
We make use of info boxes to display information on daily & cumulative cases, deaths, and vaccinations. This requiresus to perform a subset command in the server side based on the user input and update the ui with the output. The aim is to provide key statistics colour-coded at a glance. Developers can add functionality such as choosing which statistics to display.
Data Table Tab
We then want to provide users with an interactive space to query the data for themselves, by country. The DT package makes this very straightforward and quick to implement. The result is an interactive table that can be queried and sorted.
This interactive element allows for complete datasets to be loaded succinctly. Further features can be added such as user-inputted SQL commands through the sqldf package.
Figure 2: Data tables can be queried and inspected at a deeper level by the user
This requires generalised plots that update based on user input. The user can set two countries they wish to view, their date range, the feature they want to plot, and the option to include a moving average plot.
Special care must be taken to update the headings and legends to match the user inputs. These can be saved and used when compiling internal reports. Users are able to generate multiple plots without having to produce any newcode. The moving average functionality allows users to stipulate the number of days they want to average over, and a plotis produced below the regular plot. The rollmean() function found in the zoo package is used.
Figure 3: Moving average pltos allow noise to be removed and patterns to be analysed.
When dashboards are created, they are deployed locally and accessed through localhost. While this is useful for testing features but it is inflexible if the aim is to distribute it with others, since it would require them to install R and all dependencies on their local machine. Some solutions are:
Shinyapps.io allows developers to quickly deploy their dashboards on pre-existing hardware that is secured by RStudio themselves. Free-tier gives developers a limited number of active hours but up to 5 dashboards on a preset domain,which scales up through various paid tiers allowing for multiple dashboards and personalised domains.
This is an open-source solution through Amazon Web Servers or a similar cloud provider. Developers have to build up the environment themselves through Linux-supported binaries which may be advanced for some but offers flexibility and personalised control over their deployment and management of their server.
This is a collaborative environment for all R hosting from dashboards to R markdown reports through RStudio. It is a commercial licence and may be suitable for a team working collaboratively in R using RStudio's suite that wish to deploy their software and receive the necessary support from RStudio.
The dashboard can be accessed here:
References and Further Reading
 Hosting and deployment.
 Learn shiny.
 Open source & professional software for data science teams, May 2021.
 Diana Beltekian Hannah Ritchie, Esteban Ortiz-Ospina. Coronavirus pandemic (covid-19). Our World in Data, 2020.
This course guides you through the data science pipeline as it applies to the R programming language. We believe in building programming skills from the ground up so we dive into the basics of R before exploring R's wealth of packages. We establish programming fundamentals and building blocks which will allow you to tackle R projects logically and apply these techniques to other programming languages you may wish to learn in the future
The course is structured as follows
- Getting Started: An introduction to R, data science, and the tools used
- Problem Specification: How R can help your business and simple use cases
- Data Collection: Programming fundamentals, using R's data structures, importing external data, and using R's internal datasets
- Data Management: Conditionals, loops, and creating functions
- Model Building: Using mathematical and statistical models in R, including built-in statistical distributions and functions
- Data Visualisation: Various methods of visualising data in R
Data Science Vacancy 1
Your key responsibility will be to lead the Data Science team to not only transform this data into actionable insights for our customers but also develop the intelligent, data driven insurance solution into the market.
Our client; a rapidly growing mobile tech company is recruiting.
- Bachelor’s degree in IT, Computer Science or equivalent with a background in
- General Insurance sector (Motor insurance would be ideal) and at least 3-4 years of
- experience as a Data Scientist
- Worked on large open source datasets to perform data cleansing, wrangling and
- Previous experience of EDA and identifying patterns in data using Python/R
- Strong fundamentals of statistics and knowledge of applying machine learning models
- according to the business scenario
- Experience of developing machine learning and deep learning algorithms in Python
- Experience of deploying machine learning models in cloud (AWS preferred)
- Experience with NoSQL database
- Preference will be given to candidates with background particularly in Motor as well
- as commercial fleet Insurance
- Candidates must have a keen sense of pattern detection and anomaly detection to
- identify patterns in data
- Ability to apply the principles of AI, database systems, human/computer interaction,
- and numerical analysis would be a plus
Contact us via email with your CV and a cover letter if you are interested.
Data Science Vacancy 2
Contact us via email with your CV and a cover letter if you are interested.
FOUNDATIONS IN R FOR ACTUARIES
Introduction of the data science pipeline. Fundamentals of R . Building your first data science model. Data management tools within R Regression analysis and statistical packages. Powerful visualisation.
FOUNDATIONS IN PYTHON FOR ACTUARIES
Learn the fundamentals of Python through interactive Jupyter Notebooks. Discover data management tools & techniques using pandas and NumPy. Explore regression analysis, model building, validation & visualisations
PRACTICAL DATA SCIENCE TASTER LESSONS IN R AND PYTHON
This is a taster to R & Python for beginners. Through a combination of webinars & interactive notebooks, users are able to get a taste of R and Python with the use of a practical example
VISUALISATIONS DEEP DIVE IN R
Apply visualisation techniques with the use of R and some of its main visualisation packages to present frequency & severity for natural disasters and COVID experience, to interpret the impact, patterns and trends
MORTALITY EXPERIENCE ANALYSIS IN R
An end-to-end walkthrough of mortality modelling using machine learning techniques covering cleaning the data, creating visualisations to understand drivers of mortality. PREREQUISITE: Foundations in R
LAPSE EXPERIENCE ANALYSIS
Access to an end-to-end walk through investigating lapse rates in a life insurance context incl cleaning the data, preliminary visualisations, feature engineering, fitting models and reporting
MOTOR PRICING AND RESERVING IN PYTHON
We utilise data science techniques to analyse motor claims By fitting advanced data science models we aim to predict the frequency of these claims to assess risk, in the context of pricing and reserving
VISUALISATIONS CONCEPTUAL INTRODUCTION
Data Science Conceptual Overview of Visualisations. Includes: the benefits of data visualisation, overview of types of visualisations; tools, best practice techniques and much more
IFRS 17 INTRODUCTION: HANDS ON TRAINING
Welcome to the introduction to IFRS 17; hands-on training using practical examples
For information and to sign-up,
please click here
For corporate subscriptions email: email@example.com
By signing up, you also get access to our rich resource library containing webinars and papers on topics around data science, reporting, and actuarial applications.
You will also have the opportunity to network and share ideas with other professionals through our topical forum discussions.
We are still hard at work putting together future courses and case studies to be released. If you have any specific training suggestions or needs; including access to our corporate bundles please let us know here.
Listen to our recent webinars on-demand here: Exploring Ethics & AI within Financial Service | Interpretable Machine Learning
We are the home of actuarial technology for insight, training and networking
Actuartech provides leading insights and research on the topic of actuarial technology.
We take technology and data science beyond the theoretical and can uncover, and explain, in simple to understand language the impact it will have and the problems we are trying to solve and ultimately what value it will bring to the organisation. We cut through the noise and bring you the simple, easy to understand insight and practical examples.
Find out more about sponsorship opportunities with Actuartech for upcoming podcasts, videos, and events
Actuartech wants to hear from you!
We would love to hear your thoughts on our newsletter and the articles featured and any suggestions you may have for next month.
Our Facebook page is live - please take a look and give us a Like!
We look forward to hearing from you and thank you for reading.