Usually, this process requires multiple statistical techniques ranging from data mining, predictive modeling, and machine learning so it can be properly set up.
With the proper model setup, it is possible to generate risk assessments, and multiple behavior predictions based on historical analysis.
In short, it can be summarized as using a highly capable and trained AI to look at past events, trace patterns, and estimate/ predict how it will evolve.
Due to the large amount of data an AI can process simultaneously, it is possible to get better and more detailed predictions only compared to human intuition, with the difference that with predictive analytics we can trace the derivative process.