How design a ranking algorithsm
Posted on 2011-09-14
I have a few many factors to consider when designing algorithms for ordering thousands of hit records.
1. The number of hits on the urls;
2. The number of the urls being shared among friends;
3. The number of recommendations on the urls;
4. The number of not-recommendations on the urls.
The score based on the above four factors should be converged into (0, 1]. In addition, there is a weight for each of the factors, say 20%, 30%, 40% and 10%.
Note please that the number of hits mostly is much larger than the number of each of the rest factors. I need to remove such skews in the score calculation.