Data mining to improve biosecurity risk profiling
This project will conduct a series of case studies to test and demonstrate the value of data mining for risk profiling, and will determine how to incorporate these techniques in operational practices. It will:
- Draw together the different profiling requirements across the department and investigate which techniques are appropriate for which pathways
- Search for commonalities and clear differences in the approaches required.
- Develop repeatable analysis algorithms to enable statistical profiling for each case study.
- Apply these techniques to the different pathways and data sources in the department in a series of case studies.
- Determine how the insights from profiling can be communicated to stakeholders and thereby improve compliance. Develop and then simplify data extraction, preprocessing and transformation techniques so that they can be incorporated into DA IT systems and business practices.
The suite of case studies include:
- Spatial referencing of postal addresses (geocoding) to augment compliance data with census data from the Australian Bureau of Statistics (Chris Woodland)
- Generalised Pattern Analysis—analyse traveller-related data to identify better risk indicators for risk-profiling passengers (Kathleen Quan)
- Analysing patterns of import broker activity to identify and predict non-compliance (Stephen Richardson)
- Transforming AIMS data to define import units for risk-return analysis.
- Estimating compliance when evidence is scarce or incomplete (Nianjun Liu)
- Broad patterns of compliance in imported cargo.
- Predicting the frequency of hitchhiker pests based on season, cargo-type and other risk factors (Jamie Brown)