Dear Shahab,

Could you please help by briefly providing a few more details? Which algorithms/parameters/metrics have you used on what kind of data? What did your training and test sets consist of?

Kind regards,

Georgios

Have a look at the domain of "Multicriteria Decision Analysis" (MCDA)

You should decide an objective function, or multi-value objective function. Then you should select the ranking system that get the best value of the objective function.

Dear Georgios Drakakis

I have recommended books with the use of self proposed rank aggregation algorithm, by using three different techniques, I have three different ranking, I would like to check which is the best one?

1. books are recommended with the help of user-feedback having some calculation .

2. books are recommended based on link mining by taking ranks of some standard ranker.

regards

Shahab

Dear Christine Decaestecker

I have used OWA (MCDM: Multi Criteria Decision Making problem), and other techniques (self proposed and designed based on MCDM), and different techniques gives different ranking, now how to compare these?

Regards

Shahab

Dear Ahmed Hassan Yousef

Thanks for your ideas, will try for it and sure let you know if anything works.

Regards

Shahab

I'm not sure I understood your problem 100% but it sounds like you may be able to aggregate using the Kemeny rule. Let me know if this works.

Cheers

Georgios

Dear georgios

I am not familiar with Kemeny rule, however I will try it and will surely let you know the updates.

Sincerely

Shahab

Different approaches consider diferent criteria or different criteria weights (preference agregation models) towards a final comparative value. That's why you get differences. So you might be looking for a neutral (non-parametric) technique, that assumes nothing on criteria weights (importance) and lets the model decide what's best, from all your ranking possibilities. DEA (data envelopment analysis) is one such ranking technique. You will find a lot of information about this, even in RG. You just have to feed DEA with all the criteria you find relevant (from every ranking approach), provide the values for each criteria and DEA will rank the alternatives (decision making units) by comparing their performance with a linear combination of other alternatives values in every criteria. This way you can let the model find a new "best ranking" from a sort of combination between the current rankings. Ranking criteria should be consistent between them (complete, non-overlapping, preferential independence, etc.. - this comes from MCDM).

Good luck! 

Hi, you can give weight distribution for each approaches. You can simply add all the scores of each approaches, then you can re-rank it again and get the best solution. So, first you need to specify the appropriate weight for each approach.

the different quantifier of OWA is used to get different values, based upon those values ranking is done. Now still struggling in getting final ranking??