0

I'm trying to make the best decision for assignment of books (some of the books have multiple authors that belong to different organizations). Only one author can be represented by a single book. I have two variables that I'm calculating, namely the Personal Significance Index (PSI) and Organization Significance Index (OSI). I would like to base my decision on those two factors but with varying significance of each factor.

I would like to assess the "quality of the distribution of books" by changing the impact of PSI and OSI in the decision process.

As an example lets say that I have an Author and I want to represent him by his book, taking into consideration that the book has multiple authors, and that the other authors would like to be represented by the same book. I have calculated the PSI of the book for the given author, and a OSI for the given book, and I would like to make the decision.

As an example:

Let's consider that the PSI is 0.9 and the OSI is 0.4:

In the decision process I cross-check the PSI of each author and I would like to assign a book to the author with the highest PSI, but also taking into consideration that I would represent an institution (OSI).

Is it possible to fuzzy model the decision so that I can base my decision more on one index and less on other one? For example:

PSI (0.9) 30% of impact on the decision OSI (0.4) 70% of impact on the decision

Perhaps making multiple passes with changing the "impact" percentage and assess the quality of distribution?

Jay Kominek
  • 8,674
  • 1
  • 34
  • 51
Composer
  • 245
  • 1
  • 4
  • 17
  • I think this question would be much better suited for maths or algorithms. Personally as programmer i don't understand the question at all. – bdecaf May 23 '15 at 11:53
  • @bdecaf Thank you for the suggestion. Can you please suggest me the proper forum and/or keywords? – Composer May 23 '15 at 12:17
  • I'm just guessing but maybe http://math.stackexchange.com or http://programmers.stackexchange.com – bdecaf May 23 '15 at 16:19
  • You could just pick the first author listed. They are almost always the primary author. – Robert Harvey May 23 '15 at 16:38

1 Answers1

0

You could just linearly weight the "Significance Indicies". If PSI is 0.9 and OSI is 0.4, but you use a 30% and 70% weighting (respectively), multiply 0.9 by 0.3 and 0.4 by 0.7, to get 0.27 and 0.28. Then decide based on the relative magnitudes. (You'd choose to assign by Organization in this case.)

Or, you could make one of the weights positive and the other negative, and then sum the values, and make your decision based on the sign. If you repeat the above example, using negative weights for the PSI, then you end up computing 0.9*(-0.3) + 0.4*0.7) = 0.01, which is (unsurprisingly) positive, so you still assign the book by Organization.

At that point, you've constructed a Linear classifier. If you find yourself unable to come up with weights that you like, but you can look at books on a case by case basis, and label them, there are a number of common algorithms that could determine the "best" weights for you (for varying definitions of "best").

Jay Kominek
  • 8,674
  • 1
  • 34
  • 51