Web based systems for actuarial work

Overview

During the past few decades, the actuarial models have become more dynamic. Professionals are using stochastic models to process thousands of scenarios and sensitivity testing has become an integral part of the actuarial modelling process. As a result, the actuarial departments of various insurance companies invested heavily in clusters of compute intensive servers with 500 or more computing cores. Such setup requires constant monitoring by skilled IT staff and the cost of computing subsequently increased. [1]

The increased cost of computing hardware has made the companies look for alternative solutions and cloud computing is becoming a favourable choice. Companies like Amazon Web Services (AWS), Google and Microsoft offers affordable solutions for GPU and parallel computing, which has motivated actuaries to start using cloud services to do actuarial modelling. In recent times, the quantity of data has increased enormously. This has also forced actuarial professionals to look for software solutions which can handle large amount of data with the capability of machine learning and artificial intelligence. [2] Cloud computing companies package such services based on type of servers, such as Web, Database, Load balancer, data streaming, storage etc. Actuaries also need software applications to handle the data and run their models. Many professionals find it suitable to choose open source applications such as R and Python for the modelling requirements. Such open source solutions are catering to a larger audience and have made a significant impact for machine learning and data analysis work done by professionals. There are many libraries made by community contributors to do specific actuarial work in R and Python and they are often sufficient for the actuarial modelling work. Also, there are many commercial solutions available in the market to perform targeted actuarial work. Insurance companies invest heavily in acquisition and training their actuarial staff on such applications to create capability within their departments. However, most of these commercial applications are offered as an on-demand cloud solution and requires intensive cloud computing and the learning curve is steep.

Cloud Computing

In a recent experiment by Milliman employees, an actuarial modelling use case was taken and they were able to cut a three-month machine learning exploration project down to just under four days using a mixture of open source tools and the Microsoft Azure cloud. Machine learning is spreading quickly across many industries and is showing promising results for making better predictions and automating manual tasks. [3] The saving of time on conducting mammoth modelling tasks is significant as it allows us to spend more time on the evaluation and take informed business decisions. Moreover, most of the cloud computing service providers offer pay-as-you-go option to store data and run applications.

There is a large gap between the available cloud based solutions for actuarial modelling and the requirements. With the increased penetration of advanced mobile devices in our lives, easy access to availability of high speed internet and modern technology, it is essential that a cloud based application should harness the power of cloud computing. An ideal application should have a web interface which allows the actuarial professional to upload their datasets, choose modelling requirements and the result is displayed on the user’s device after all the calculations are performed on the host server.

Cloud computing is providing actuaries with the ability to focus on what matters most and avoid undifferentiated work like procurement, maintenance, and capacity planning. As cloud computing has grown in popularity, several different models and deployment strategies have emerged to help meet specific needs of actuarial work. Each type of cloud service, and deployment method, provides different levels of control, flexibility, and management. There are many options for an actuarial department to run their modelling work on the cloud, such as:

Infrastructure as a Service (IaaS): contains the basic building blocks for cloud IT and typically provide access to networking features, computers (virtual or on dedicated hardware), and data storage space. IaaS provides the highest level of flexibility and management control over the required IT resources for actuarial modelling and is most similar to running a virtual data centre.

Platform as a Service (PaaS): remove the need for actuarial departments to manage the underlying infrastructure (usually hardware and operating systems) and allows to focus on the deployment and management of actuarial models. This helps actuaries to be more efficient as they don’t need to worry about resource procurement, capacity planning, software maintenance, patching, or any of the other undifferentiated heavy lifting involved in running the models.

Software as a Service (SaaS): is a completed product that is run and managed by the vendor. With a SaaS application, the actuarial department do not have to think about how the service is maintained or how the underlying infrastructure is managed; they only need to think about how they will use that particular piece of software. A common example of a SaaS application is Moody’s Analytics AXIS actuarial system is widely used by life insurers, reinsurers, and consulting firms for pricing, reserving, asset and liability management (ALM), financial modelling, capital calculations, and hedging.

Use Cases

Actuarial work is characterised by many complex calculations, depending on the use of a great number of parameters. By varying the values of such parameters and re-running these calculations, we get the result as an even greater number of simulations. The results of these simulations are collated and allows the actuaries to be sure of their estimates with certain assumptions.

Electronic spreadsheets are often used to make such calculations because they are easy to learn, very intuitive and relatively easy to maintain. However, spreadsheets do present some problems. Their ease of use promotes the rapid development of multiple methods with very small differences. By doing this, lack of uniformity may occur for no reason. It is very easy to forget to change a formula that occurs in a corner of multiple spreadsheets, which are all part of a single method. Saving each version of such a multi-sheet simulation can lead to thousands of stored spreadsheets. Comparing those spreadsheets to find minute differences turns into a very difficult and time consuming task. Moreover, spreadsheet calculations tend to mix the data (a database) and the logic (a program) into a document (a report); it becomes impossible to manage these entities in an audit-able structure.

In addition, these actuaries often find themselves working long hours and against short deadlines. A high level of concentration must be maintained because they must keep track of their simulations in order to compare their results. They are obliged to present a clear but concise record of the results they have obtained. There are lot of redundant works that an actuary performs based on the sensitivity analysis and these results are most often linked together in a complex network to obtain a consolidated vision. They share primary indemnity data, hypotheses as well as intermediate and consolidated results. Therefore, a web based actuarial valuation system can present an opportunity to consolidate the results and keep a track of many versions of the results. Such system can also help to do the iterative nature of the actuarial work with more confidence.

Open Source Applications

As we all know that, R and Python are the most popular programming languages for analytical work. The good news is that there are some useful packages available in both the platforms and they can significantly reduce the time needed for modelling work:

R: actuar, lifecontingencies, ELT, DCL etc.

Python: lifelib, Numpy, Pandas, SciPy etc.

All such packages are free to use and are not compute intensive. They can be easily installed on a virtual instance of any cloud computing service. Many service providers also offer training manuals and tutorials to setup R and Python environment. It is surely an easy and affordable approach to setup a cloud instance of these popular programming languages and start the modelling work without much of a hassle. These instances are scalable and are easy to destroy/replace when the work is over.

The challenge to use the open source applications is to build own models and for that one has to be proficient in the use of R/Python. Everything is not out of the box as these programmes are primarily meant for data sciences work and not for actuarial sciences. Therefore, additional fragments of codes are mostly required to complete the task and generator the reports in the required format.

Commercial Applications

Although, there are many actuarial applications available these days which are capable of handling actuarial work through a cloud based system. Such applications are not easy to acquire and deployed due to the complex nature of actuarial valuations. The cost to deploy commercial cloud solution has an upfront cost and also a recurring cost to manage the instance. Also, customisations of these applications is not generally possible and only few big organisations can afford a custom workflow in such applications. As these are vendor managed instances, there is less control over data. Regulations also play an important role in how insurance companies manage their data. Usually it is not allowed to host any such data outside the geographical boundaries of the country. In the beginning most of the data centres, where vendors were managing the applications were based outside India and therefore they were forced to either have in-house infrastructure to comply with the regulations or outsource to a local data centre which has the facility to host data securely. In past 2 years, the situation has changed significantly. Many big cloud computing companies, including Amazon Web Services, Microsoft and Google have started their data centres in India making it possible for actuaries to use their services and host their cloud applications and data within India.

Conclusion

With the affordable and on demand cloud computing available in the present times, it makes a lot of sense to develop custom workflows which are capable of performing actuarial modelling work through a point and click system with few useful options for targeted actuarial work.

References

[1] "The Digital Insurer," 10 March 2019. [Online]. Available: https://www.the-digitalinsurer.com/actuarial-models-meet-the-cloud/. [Accessed 23 March 2019].

[2] "Why Actuaries Should Start Paying Attention to Python by Andrew Webster," 13 June 2018. [Online]. Available: https://blog.actexmadriver.com/2018/06/13/why-actuariesshould-start-paying-attention-to-python. [Accessed 24 March 2019].

[3] "Parallel cloud computing: Making massive actuarial risk analysis possible" 2 May 2018.[Online]. Available: https://us.milliman.com/insight/2018/Parallel-cloud-computing-Making-massive-actuarial-risk-analysis-possible/. [Accessed 24 March 2019].