The underlying stochastic nature of the Solvency II regulations has introduced significant computational challenges if the required calculations are to be performed correctly, within practical timescales. The ubiquitous approach of using a larger volume of more up-to-date hardware to run industry-standard software cannot solve the problem. We aim to show that, using a combination of algorithmic improvements, serial optimisations and multi-care parallelism, it is possible to perform the required calculations in timescales of the order of a few hours.
For profitability calculations, we have already obtained a speedup of roughly 200,000 times over commercial packages; after removing effects of parallelism, we achieved a single-threaded speedup of roughly 4,000 times. One of the key factors in achieving this performance gain was to calculate reserves using a far more efficient algorithm which is based on the survival states of the lives involved, rather than on the type of policy. We are now at the point where we expect to perform a full brute force calculation of liabilities on a portfolio of the order of a million policies in under an hour. Therefore we would like to extend the research to include other tasks which are currently beyond computational contemplation; however, we would like some guidance in identifying what those tasks might be.
Registration - 08.30
Start - 09.00
Close - 10.00
Location
Address
Moody's Analytics Ltd,
7 Exchange Crescent,
Edinburgh,
EH3 8RD
Nearest Public Transport
Haymarket