Machine Learning to identify Ad Servers in Clickstream

Being able to sift through the clickstream of thousands of people is interesting in the search of insight is a fun task.

However clickstream is fill of noise:

• Ad servers;
• Iframes
• Analtyics trackers (adsence, commscore etc...)

Ad servers are particularly annoying - they are shape shifters - continually adding changing both domains and subdomains.

However - noise has patterns:

• Referrers
• Redirect Codes
• HTTP Headers
• In-discriminant appearances across the web.

These are the sort of patterns that machine learning could eat for breakfast!