CODEWITHSUNDEEP
machine-learningunsupervised-learningboosting
Rohan Thakkar
Rohan Thakkar

Reputation: 1

Gradient Boosting Tree for feature engineering? (Rank)

I am working on a project where I have clicks/dismisses on web pages and I want to rank the web pages accordingly. I initially did this by writing a simple equation for calculating relevance score but I am hearing that random boosting tree can help with this. How exactly can I engineer the 'rank' feature for pages using random boosting tree though? I have only little experience in ML and its in supervised learning where you have some output in training.

Upvotes: 0

Views: 277

Answers (1)

abhiieor
abhiieor

Reputation: 3554

This could be a classification problem. Create a new column for your data called click; for the click instances give it value 1 else 0. Now fit a boosting (or any other machine learning) model. In classification problem you get probability score [0,1] as output. Naive approach to convert this probability scores to actual click/dismiss is simple rounding. So in effect 0-0.5 probability score records will go into dismiss and 0.5-1 probability score records will go into click.

Having said all this the most difficult step of all this will be actually fitting the model which takes substantial ML, maths, domain knowledge along with ingenuity. Best of Luck.

Upvotes: 0

Related Questions