I need to predict age or age range of a user according to his/her browsing history from a news website. The only data I am able to get is USER-ID, URLs-visited, TAGs from those URLs, "which category news a person reads most" etc. My only option seems like creating this dataset somehow. But I do not know how do I decide the age range of a person. The problem is if a person is reading a politics news he could belong to an age range of 50 and above but then what if this is a high trending news and everybody is reading it. So basically how do I decide which age range people read which kind of news most?

I hope I am able to explain what my problem is. If somebody has worked on such a problem I would like to hear how you proceeded with it. Thanks.

  • $\begingroup$ In addition , we can also visit this link and can have a bit of idea about this problem. Data is not accessible but still we can see different parameters by looking at the file names. $\endgroup$ – Konark Yadav Dec 15 '17 at 7:23
  • $\begingroup$ This is exactly what I am looking for, any idea how I can access this dataset? Or a dataset similar to this would solve my problem. $\endgroup$ – Ankita Sharma Dec 15 '17 at 7:26
  • $\begingroup$ I would suggest to do unsupervised clustering and come up to the logical choices by experimentations. $\endgroup$ – Gopal Krishna Varshney Dec 15 '17 at 10:19
  • $\begingroup$ This is a classification problem.. We can have defined classes, for eg: 18-24, 25-35, 36-46, 47-60 and 60 and above. What we do not have is a behavioural dataset that can be used for training. We have users and their browsing history, what I need to do is classify one user into one of the defined age ranges. For this I need a pre defined behavioural dataset. As in, how do teenagers behave when it comes to news reading or what kind of news do oldies read? If I can come up with a set defined behavior pattern of different age groups then we might be able to create the dataset by ourselves. $\endgroup$ – Ankita Sharma Dec 15 '17 at 11:48

This problem is quite broad and I would like to add some of my suggestions. Refer to good source of Age wise browsing data, age-wise social behaviors and their preferences publicly available. Here are some datasource that you may refer and you will have to do extensive research on lot of factors to predict age based on browsing data.

Datasource 1: https://www.google.com/publicdata/directory

Datasource 2: https://www.ciser.cornell.edu/ASPs/datasource.asp

Refer to this paper : https://link.springer.com/chapter/10.1007/3-540-44934-5_3

I would definitely say that this problem doesn't seem to have any perfect solution, only thing we may do is to try our level best by various sources and implementations. Learning algorithms on public datasets on behavior, survey, social and mental age specific activities etc can help as well. I have also attached a link to a paper on similar problem. I hope this helps to some extent.


Not the answer you're looking for? Browse other questions tagged or ask your own question.