The Daily Crime Forecast (DCF) varies significantly from current hot spotting routines in that it is predictive instead of retrospective. Specifically the DCF is different in the following ways:
Spatial clustering routine: The DCF utilizes an innovative clustering methodology that:
- Can be generated automatically without user input
- Ensures that the size of clusters are fixed to allow efficient coverage (target areas are not too big or too small)
- Will not produce clusters that overlap
- Will accurately identify ‘linear’ risk areas
- Instead of providing 'hot spots' areas based on user-chosen parameters as in most spatial hot spotting routines, the DCF builds its risk areas based on available resources
Evaluates Spatial and Temporal attributes: The DCF uses a novel methodology that uses machine learning algorithms to capitalize on the spatial and temporal distribution of crime by effectively assessing the relevance of attributes within the data. Using several relevant temporal and spatial factors together, a better and more accurate prediction can be made to predicts times and locations where crime is likely to occur.
Adjusts for current temporal conditions: Instead of showing crime hot spots for the last weeks or months as in existing methodologies, the DCF ensures that it evaluates the current temporal conditions before generating a forecast. Intuitively we know that the propensity for crime will vary not only by location but also by day of week, month or even day of month. The DCF takes into account the temporal conditions of the day being predicted in generating its forecast. This allows for the DCF to output a new and optimized forecast for every day of the year.
Automation: The DCF is fully automated and can produce a daily crime forecast without user interaction or the need for pre or post analysis. The DCF will fetch the required data, then generate and output the forecasts to a secure web portal automatically for every day of the year. That means that your analysts can concentrate other projects knowing that patrol deployments are taken care of in the most efficient manner. This also means that the forecasts will also be generated on weekends and holidays.