Anomalies inform us about anything we are interested in because they can affect our actions. Therefore, it can be important for managers and administrators to quickly detect deviations. But how do you prevent an abundance of new information? And do you guarantee the quality of the anomalies that you observe in the data?
If you want to know what the deviations are from data, and you don’t want to generate false alarms, it is necessary to recognize fluctuations such as seasonal patterns and other predictable – sometimes desirable – economic conditions in your data. There are many fluctuations in data sets that are predictable. Whether we are talking about stock, logistics, food production, or people’s purchasing behavior during holidays, which can be used to create risk profiles which in turn are used in fraud prevention.
Jort de Bokx is a Master student Data Science in Engineering at TU/e. He is involved in optimizing CodeNext21’s ML algorithms that recognize seasonal patterns in data. With this, we increase the quality of our ML21 platform.
About the Author: Remco Eikhout
