Predvol, the tool the French police force use to predict areas where there is a risk of vehicles being stolen, improves police operations

In 2015, the General Administration of Data Team (AGD) within the interministérielle du numérique et du système d’information et de communication de l’État (DINSIC),[1] developed, in collaboration with the Service of Technologies and Interior Security Information Systems (ST(SI),2 a model to predict vehicle-related robberies. This collaboration has facilitated the Predvol, a tool to help with police decision-making to predict the risk of theft on a daily basis, presents a record of thefts and a classification of neighbourhoods in accordance with the type of offences that take place there.

Predvol has been optimised to be used with a touch screen and has been implemented in the French department of Oise (in the north of the country), which is especially prone to vehicle theft. More specifically, the gendarmes have applied it as a decision-making tool and have tested it in the city of Compiègne (one of the main cities of Oise) as of May 2016. Moreover, the anti-crime squad of the department’s public security board of the national police tested it in the city of Beauvais, the department’s capital.

This programme is a predictive tool[2] that uses a large number of variables and algorithms that enable the police to decide which variables are the best predictors to anticipate car theft.

After experimenting for six months, one of the conclusions has been that the attention of the operating systems focused more on visualising the crimes committed rather than the prediction of theft on a daily basis. Hence, the predictions were very effective, but only served to confirm high-risk areas that are already known to the system; on the other hand, the simple representation of incidents on a map implied improvements in the daily service. This realisation facilitated the introduction of improvements in the visualisation of offences. The visualisation of offences from the moment a report is filed enables the police to gather quality data and provide better results in police work.

For further information:

[1] The DINSIC is answerable to the prime minister within the French public administration and is in charge of coordinating the actions of administrations in terms of information.;

[2] Other predictive models: “logistic regression”, “random method”, XGBoost, Boosting, PredPol (predictive policing), evolutionary heat mapping…


Aquest apunt en català / Esta entrada en español / Post en français


Deixa un comentari