ISPs sometimes offer more performance when a customer is doing a load 
test than any other time. The authors of this paper use a 
machine-learning model to detect "fake" bandwidth using 20 days of logs.

www.researchgate.net

(PDF) Optimisation of Network Logs for Fake Bandwidth Classification 
using CNN <#>

PDF | The goal of this study is to enhance the classification accuracy 
of fake bandwidth using a CNN model, leveraging network logs collected 
in... | Find, read and cite all the research you need on ResearchGate

🔗 
https://www.researchgate.net/publication/392480826_Optimisation_of_Network_Logs_for_Fake_Bandwidth_Classification_using_CNN 
<https://www.researchgate.net/publication/392480826_Optimisation_of_Network_Logs_for_Fake_Bandwidth_Classification_using_CNN> 


  The work also looks at the possibility of using the "real" bandwidth 
to guide allocation and keep the QoS high

--dave

-- 
David Collier-Brown,         | Always do right. This will gratify
System Programmer and Author | some people and astonish the rest
davecb@spamcop.net           |              -- Mark Twain