The stream on Actuarial Software seeks to publish contributions that provide to the actuarial and insurance community valuable computing tools, with clearly explained methodology and relevant practical use cases. This offering recognises the importance of practical software toolboxes to the development of actuarial research and practice.
This stream of the AAS publishes well-documented and tested open-source software tools that address important applications in insurance, risk management and quantitative finance. We welcome implementations of methods already published in the peer-reviewed academic literature, as well as methodological innovations, where the software implementation provides a major part of the contribution. Technical contributions can be in areas including stochastic simulation, numerical methods, machine learning, computational statistics, and data science, while possible applications include life insurance, non-life insurance, pensions, health insurance, finance and investment, econometrics, insurance economics and financial risk management. All topics will be considered, as long as the contribution is clearly defined and includes clearly worked examples that can be reproduced by users of the tool.
All submissions, including articles and code, are subject to a rigorous process of peer-review.
Please note that the Actuarial Software stream does not publish pure methodological innovations in actuarial science or pure application of existing packages to actuarial problems / comparison of existing packages. The papers considered in this series should contribute both novel software tools and clear and detailed applications. Contributions that primarily consist of novel models or estimation procedures, should be submitted under the more general Original Research Papers stream of the AAS. Implementation of existing methods and models should not be an incremental refinement of an existing software solution, or simply a translation of an existing software solution from one language to another, unless there is a clearly motivated and justified need for such a solution.
Implementations should be written in languages widely used in industry and academia. The software published under the Actuarial Software stream of the AAS will be open-source and freely available in the public domain to potential users, for unrestricted use under a GNU General Public Licence (versions GPL-2 or GPL-3). This stream is committed to open-source free software, which respects users' essential freedoms, including the right to run it, to study and change it, and to redistribute copies with or without changes.
Submissions will include both code and an article. Full reproducibility is mandatory for publication and the source code is published by the AAS along with the article as supplementary material. Submissions should be consistent with the AAS Transparency and Openness Promotion Policy
Code can be in any of the following interpreted or compiled high-level languages that have wide utilisation in actuarial practice, e.g. R, Python, MATLAB/GNU Octave and Julia (any MATLAB script should also be proven to work in GNU Octave). Other languages may be considered subject to suitable reviewers of submissions being available; this may be particularly relevant for emerging trends such as InsureTech, RegTech and blockchain-based solutions. Any accompanying data used to reproduce code illustrations and examples must be in a suitable data format for such languages (e.g. .Rdata, Pickle, .mat, JSON, csv). Basic unit tests should be provided for the main components of the toolbox, wherever applicable. These should be summarised in the submitted paper under a section titled “Unit Testing and Validation”.
Articles must present the scope of the software and the underlying models or concepts such that readers can understand what the software does. Specifically, the article should address the following points:
- Clear exposition of the real-world problem that the software is addressing and its relevance to actuarial practice.
- Description of methodology, processes, algorithms, and models used, with appropriate citations; statistical/mathematical representations all methodological contributions implemented in the software.
- Overview of key functionalities and reference to related or existing software packages that implement such methods and justification for the novelty of the submitted work. This can focus on aspects of estimation accuracy, computational efficiency, enhancement of existing tools that exist or integration of multiple tools to produce value-added toolboxes.
- A comparison with other open-source implementations of similar models or procedures, where available, should highlight the capabilities of all implementations and the corresponding advantages or disadvantages.
- The general analysis workflow must be illustrated by non-trivial case studies on synthetic and real data. Synthetic data examples should be provided that produce illustrative and informative overviews of how one may utilise the toolbox. We strongly encourage real data application of the tool and, where applicable, clear documentation of benchmarking against competitive tools. Case studies should address the tool’s relative flexibility, accuracy, computational efficiency and ease of utilisation of outputs.
- Statistical choices, key settings and default settings should be explained and justified relative to the application they are intended for. Guidance on settings required for robust use of the software should be described and expectations for computing resources and run-time should be detailed.
We recognise that there is a variety of potential contributions to Actuarial Software and that that satisfying all the above requirements may in some circumstances not be applicable. If in doubt, authors are encouraged to contact the Editor-in-Chief for clarifications.
Upon acceptance, AAS publishes the article with the software alongside it, the latter made available as supplementary material (under a non-exclusive licence). We strongly encourage authors to additionally deposit the published version of their software and replication materials via a repository that provides a DOI and stable link (e.g. see GitHub’s connection with Zenodo), which should then be cited in the paper’s Data Availability Statement.
It will not be possible to update the article or software made available by the AAS as supplementary material once they are published. We encourage authors to make their up-to-date code available, post-publication via public repositories, e.g.: for R packages via the Comprehensive R Archive Network (CRAN); for python solutions through GitHub on a public repository; and for MATLAB on the MATLAB File Exchange with a project-linked GitHub repository. It remains the contributors’ duty to maintain and respond to queries and requests regarding such code distributions.
Actuarial Software submissions should be made through the AAS ScholarOne site.