Predictors: Geographical Processes

Photo by Toa Heftiba on Unsplash

Introduction

This section describes a set of predictive tools for certain geographical processes, such as crime and footfalls. The links to downloading these tools are also provided.

1. Grid-based Crime Hotspot Predictors

Crime Hotspot refers to an area with a high risk of crime. In practice, police are interested in crime hotspot in order to inform crime intervention. However, it requires the use of an analytical technique to identify hotspot areas and be able to rank them according to their level of severity. In a short-term crime intervention, the term predictive hotspot has been coined to represent hotspot types where crimes are expected to occur in a very near future, such as within one or two days. A number of analytical tools have been developed in order to model this type of hotspot - these are the types provided here. In addition, we provide an idea of the relative accuracies of these techniques.

**Self-Exciting Point Process**

**Prospective Space-time Scan Statistics**

**Prospective Kernel Density Estiamtion Method**

**Prospective Hotspotting**

2. Footfall Predictor

This tool was developed as a part of the Surf Project funded by ESRC Future Research Leaders scheme. The tool, named as Leeds Footfall Predictor, built on the work led by Dr. Nick Malleson which determined that the daily footfall rates observed in the Leeds city centre can be explained in terms of external variables, such as temperature, rainfall and holidays (See here). The research involved the application of different machine learning algorithms to a combination of these variables (predictors) and subsequently compared the accuracies of the former. It was established that the Random Forest algorithm by Breiman, 2001 is the most accurate for forecasting footfall rates. Thus, the algorithm, as implemented in R by Liaw and Wiener, 2002, was employed in the development of the Leeds Footfall Predictor. The tool was then deployed using the RShinydashboard platform.

Figures here (Developed by Monsuru Adepeju)

Figure above shows the homepage of the tool. For further details about Surf Project, please contact Dr. Nick Malleson

Dr. Monsuru Adepeju
Dr. Monsuru Adepeju
Senior Research Associate

My research interests include Inequality modelling, Spatiotemporal modelling, Police data analytics.