Short description

Forensic evidence evaluation begins with measurements on forensic traces. Nowadays, we usually get loads of measurements from one trace. For example over 1000 frequency-bands for speech analysis, highly detailed photographs for face comparison or over 40 elements for glass elemental composition. So we are often rich in data, but unfortunately we are sample poor: we only have one telephone recording or one photo per perpetrator, or one glass fragment on a piece of clothing.
This calls for the need of dimension reduction techniques in order to implement feasible statistical techniques for evidence evaluation. At the moment, dimension reduction techniques are often applied on an ad-hoc basis, and there is a need for a systematic overview in order to more deliberately apply them. The student is asked to provide such an overview in this literature thesis, with pros and cons for application for forensic likelihood ratios.

References

1. https://towardsdatascience.com/11-dimensionality-reduction-techniques-you-should-know-in-2021-dcb9500d388b
2. https://www.analyticsvidhya.com/blog/2015/07/dimension-reduction-methods/
3. Interpreting Evidence, B. Robertson, G.A. Vignaux, C.E.H. Berger, 2nd ed. Wiley

Required / recommended expertise

• Good understanding of statistical methods is required.

Information

Institute / Company: Netherlands Forensic Institute
Country: The Netherlands
Supervisor: Peter Vergeer
UvA Co-assessor: Marjan Sjerps