popularity - Algorithm for most popular posts based on likes, shares and views -
this question has answer here:
- popularity algorithm 5 answers
i working on website have gazillions of stories. stories in formats: texts, videos, photos , other multimedia elements. stories can filtered on various basis of "new" contain latest stories first, "featured" stories marked featured manually , "popular" need come algorithm.
so far doing taking average of facebook likes, number of shares (including both facebook, twitter or other shares) , number of views. doesn't me. because giving equal weight-age 3 metrics doesn't sound genuine reasons social spamming etc.
looking forward algorithms rank popularity of stories.
----addition-----
popularity algorithm discusses algorithm based on "likes" , algorithm based on categorize results in categories of timestamps: popular on day, week , month. whereas this has answer answers query not because metrics assumed there. looking exact metric genuine explanation. eg "facebook *2", explanation of why *2 facebook. hope not duplicating now!
i'd suggest trying use regression algorithm. used linear regression, if model not fit - feel free explore others.
- first, determine features of each story. features likes, tweets, shares, views, .... i'd add boolean indicator (variable can values 0 or 1 only) each of types (video/photo/...).
- next, create training-set - set of stories (or other human experts) have given score to.
- now, using these features , training set - use regression algorithm create model best fits features have examples scored.1
- after have model - can use give score other documents.
regarding spammers detection - try anomality detection algorithms
(1) actually, step 2 , 3 can done - using active regression techniques - in active regression, learner (algorithm) asks examples make algorithm learn fast can. experiments palice performing active regression algorithm.
Comments
Post a Comment