|
To rank the art ArtEnthu have used the Bayesian approach. ArtEnthu takes into account the number of viewers rating the art and the ratings they have provided. It does not simply makes the average of various ratings. The idea behind a bayesian average is that it normalizes the ratings by pushing them toward the average collection rating, and it does that more for items with fewer ratings than those with more ratings. This approach makes sure that a new artwork with one or fewer vote does not immediately jumps to first place. It tries to create a balance between the number of ratings given to an art work rather than just few ratings that are high. ArtEnhtu also gives more weightage to the ratings of registered users than a non-registered user.
For Example, there are two different arts available for rating, say Art1 and Art2. Art1 has been rated by 20 users receiving a mix range of ratings from 'good' to 'excellent'. On the other hand Art2 gets 'excellent' rating only from two users.
In average rating system Art2 will get much higher average in comparison to Art1. Bayesian approach makes sure that Art1 and Art2 both when compared gets fair average by making Art1 a better art, because it accounts for the number of ratings given too. Bayesian method compares each art with the average of collection of arts. For an art to affect the Bayesian average it would require minimum number of ratings to influence the average.
Reference: Collective Choice
Improving the ranking algoithm is the continous process and there are many more factors which we are considered, few being user profile, favourites, comments, etc... from our experiences of continous online competition we have gained more insight to voting and with new ideas to improve the system. We would continue to do so.
|
