Lead Supervisor: Dr Richard Allmendinger (University of Manchester)
Other Supervisors: Dr Yu-Wang Chen
Contact Email: firstname.lastname@example.org
Partners: University of Manchester
External Partners: Forensic Testing Service Ltd
Start Date: October 2018
An oversimplification of the interpretation of analytical results in forensic testing can frequently lead to inconclusive or misleading evidence and thus miscarriages of justice. A more robust approach would take a holistic view on this based on a range of individual pieces of information including data from questionnaires and physically obtained measurements of different sample types (e.g., hair strands, blood, urine and nails). Presently, all this information is reviewed and assessed manually in terms of their likely impact on the final conclusion. This is a time-consuming and expensive process that requires improvement urgently.
The overarching aim of this project is to develop an automated data analytics tool that analyses all of the available information and then makes informative and explainable recommendations on a case by case basis. Under the hood, the tool will comprise a combination of techniques geared to first identify crucial features in the available data (feature selection), then derive explainable and statistically significant correlations in the data (via a combination of supervised and unsupervised learning methods), and finally present the data in an intuitive and informative way to make sound decisions. The research questions would be refined by the successful applicant might look something like:
- Can we produce a model which can predict accurately whether a person has taken a given drug?
- Can we embed that model in a data analytics tool?
- Can we develop such a tool in a way that produces explainable results?
The project will therefore involve a mixture of data analytics of complex data, applied machine learning and traditional social science and would suit a student with an appetite for interdisciplinary work on a socially important problem.
Reference number MN23