social networking - Scoring algorithms: how to convert the number & % of "Likes" & "Dislikes" into a single score? -


i have website users can "like" , "dislike" items.

so each item, have data such total number of "likes" , % of total votes "likes".

i'd calculate single score show users. using % wouldn't work because though item_a might have 90% of "likes" while item_b might have 80% of "likes", item_b should still rank in front of item_a if item_b has 10,000 total votes while item_a has 1,000 total votes.

likewise using total "likes" wouldn't work because while item might have large number of "likes" shouldn't ranked high if % of "likes" low.

what algorithm create single score out of data above?

ideally score should "meaningful" or "normalized" in way. example if go imdb , see movie has score of 8/10, i'd know movie. on other hand if see score of 1,370 wouldn't know if or bad.

there's couple of articles on how reddit sort of ranking here, , here. in nutshell, rank posts lower end of 90% confidence interval of scores. entries fewer votes have larger confidence intervals, , hence tend rank lower entries more votes same average.


Comments

Popular posts from this blog

asp.net - repeatedly call AddImageUrl(url) to assemble pdf document -

java - Android recognize cell phone with keyboard or not? -

iphone - How would you achieve a LED Scrolling effect? -