Break-ins do not just happen randomly anywhere at any time. The time of year or day, location and population density also play a role. In densely populated areas, machine learning techniques can be used to recognize patterns arising from break-in statistics, with the break-in risk then predicted for specific locations. Through this, the police identify so-called hotspots and accordingly deploy patrols there.
However, computers require sufficient data in order to recognize such patterns. In rural, sparsely populated areas, there is often a lack of data. Researchers at the Swiss Federal Institute of Technology in Zurich (ETH) have now developed a machine learning-based method with which precise break-in forecasts can now be made even in such sparsely populated areas.
For this, the researchers fed information into algorithms and at the end of this process combined the analysis of various algorithms. While feeding the information into the algorithms, they processed the data set: “Data units without break-ins were randomly removed until the same number of units with burglaries as units without was achieved”, as explained by the ETH in a press release. This statistical method is known as “Random Undersampling”.
ETH researcher Cristina Kadar commented: “With imbalanced data, the method achieves at least equally good and, in some cases, better hit rates than conventional methods in urban areas, where the data is denser and more evenly distributed”.
The findings are useful first and foremost for the police. According to the ETH, the method can also be used to predict other risks. For example, health risks and the probability of emergency calls to the ambulance service are cited.