ROUSSEAU Pauline

Identification of weak signals in massive data sets :

The objective of the thesis is to propose new algorithms, methods or techniques for the early identification of warning signals in varied and multi-source data, whether these data are private, public online, or mixed, private or public. During this thesis, we will focus on early warning signals of events or related to security issues, and the associated data will be analyzed in order to understand correlations and understand correlations and identify the most characteristic signals.

Law enforcement currently collects, both online and off-line, data from many different sources. from many different sources, both online and off-line. In general, the data is related to past actions that are analyzed after the fact. Although such analyses produce useful statistics, these results are rarely used. useful statistics, these results are rarely used for the anticipation and prevention of crimes or crimes or other illegal acts. Any complaint or testimony can be considered as an event, and different events can then be event, and different events can then be correlated in order to provide an explanation for a an explanation for a phenomenon. For example, different crimes in a neighborhood may be For example, different crimes in a neighborhood can be linked to the same group of people. The problem then is that such such correlations are often made “by hand”, on a case by case basis, and when it is too late (since the series of events has passed). Moreover, these correlations are established on facts, which implies a delay between the act itself and its declaration to the authorities. Having a method, judiciously applied, to detect such series of events as early as of such series of events and their correlations would allow for early action to prevent the which could prevent the development of malicious acts.