Beginning Anomaly Detection Using

Figure 8-3.  Roaming

Chapter 8   praCtiCal Use Cases of anomaly DeteCtion

---

301

If we know the various metrics of the cell phone towers and the associated devices 

at some period of time and for a long duration, along with any kind of information 

we have on the typical nature of activity around the towers in terms of whether there 

were concerts or games in the vicinity or a major event is expected in the vicinity of 

the cellular towers, we can use a time series as a basis to represent all such activity and 

subsequently use TCN or LSTM algorithms to detect anomalies pertaining to the major 

events because they have a temporal dependency. This will help in looking at how 

these services are being used and how effective the service is for the particular cell 

phone towers.

The cell phone companies now have a way of understanding whether certain hours 

need to be upgraded or more towers need to be built. For instance, if major office 

buildings are being built near a particular tower, using data on the time series of all the 

towers owned by the cellular network, it is possible to detect anomalies in other parts of 

the network and apply the principles to the tower that is probably going to be impacted 

by the newly constructed office buildings (which will add thousands of cell phone 

connections and could cause overloading on the tower and affect how the tower will be 

used in the near future).