Apologies in advance if this question is broad or basic for data science community.
Given:
Dataset containing thousands of lines with Apache HTTP Server log file produced in Common Log Format (CLF) from a library search engine:
127.0.0.1 - frank [10/Oct/2000:13:55:36 -0700] "GET /apache_pb.gif HTTP/1.0" 200 2326 "http://www.example.com/start.html" "Mozilla/4.08 [en] (Win98; I ;Nav)" 893
with the follwing pattern:
- IP address of the client (remote host)
- user_id of the person requesting the document as determined by HTTP authentication
- time that the request was received
- request line from the client is given in double quotes
- HTTP status code
- size of the object returned to the client
- Referer (IMPORTANT)
- User-Agent HTTP request header
- Session_ID
Goal:
Design a recommender system using user-user collaborative filtering.
The challenge, however, is that user_id of the users are already removed (anonymized) due to GDPR rules and privacy issues. So I am given thousands of lines of similar to this sample instead:
127.0.0.1 - - [10/Oct/2000:13:55:36 -0700] "GET /apache_pb.gif HTTP/1.0" 200 2326 "http://www.example.com/start.html" "Mozilla/4.08 [en] (Win98; I ;Nav)" 893
Questions:
Due to importance of users in user-user collaborative filtering in recommendation systems, does this idea/approach seem feasible from data science perspective? Or it seems almost impossible and I should rearrange my approach to content-based filtering or something else?
I wonder if using IP address can be somehow an alternative to user_id perhaps in this context?
My current understanding of IP address is that a user in city X has still same IP address of another user in city X regardless of its distance (within ~10 KM depending on the city of course). Therefore IP cannot evidently be used to identify a user and corresponding behavior. Or do I have a wrong understanding?
Cheers,
