This project uses Machine Learning (ML) to discover any structures and patterns in large datasets of domestic abuse incidents, victims and suspects. Specifically, we use unsupervised ML techniques to discover clusters of victims and perpetrators with common characteristics and behaviors.
Background
Domestic abuse (DA) is a national priority; it is estimated that one in five adults experience it in their lifetime. It has a far-reaching and lasting impact on victims and families, with adverse outcomes going well beyond physical injuries, mental health trauma and poor emotional wellbeing. On a community-level, DA weakens social cohesion and damages trust. Its economic and social cost is estimated at £66 billion per year, highlighting the significant burden of DA on public finances. DA is not just “a family matter”, but rather a serious social problem with devastating, widespread consequences.
Research suggests that families experiencing DA are much more likely to also face food insecurity, especially in situations where an individual controls household finances or denies access to resources making it difficult for victims and families to have enough food. In situations where financial control is not an issue, but poverty is, food shortages can create stress and tension in the household which can lead to conflict and violence.
Project Aims
The project uses Machine Learning (ML) to discover structures and patterns in large datasets of DA victims and suspects in selected police areas in the East of England. Specifically, unsupervised ML techniques will be used to analyse non-identifiable data and discover categories of incidents, victims, and perpetrators. There are plans to link data with aggregated geographical and socio-economic indictors. By bringing this data together, we will be able to understand the role of contextual variables such as health data, care data, food insecurity, nutrition, and poverty levels. We think that these social and economic factors both contribute to vulnerability and risk of DA and reflect the prevalence and systemic impact of domestic violence. When brought together, these elements show a more complete, system-level picture of the challenges related to DA, thus, helping to inform local commissioning for improved support and targeted service delivery.
Project Activity
- To use Machine Learning to analyse police datasets: our goal is to identify patterns and clusters of DA incidents, victims and perpetrators. This can help with resource allocation and identifying under-addressed areas.
- To explore links between police datasets and socio-economic indicators to understand the connection between DA and the wider determinants of health. To do this, we will analyse factors such as health data, care data, food insecurity, and poverty rates in each area.
Anticipated or actual outputs
We are working closely with two regional police forces. Our goal is to co-design the analysis with them and share our findings to help identify local needs, inform commissioning and allocation of resources. We think our project will offer insights which can support tailored public health interventions and more effective service delivery.
Who is involved?
- Alejandro Quiroz Flores, PI, University of Hertfordshire
- Losif Mporas, (Co-I) – University of Hertfordshire
- Emilia Tylenda, (Co-I) – University of Hertfordshire
Contact
Alejandro Quiroz Flores, a.quiroz-flores@herts.ac.uk
