ahpy 1.1

A Python implementation of the Analytic Hierarchy Process

AHPy

AHPy is an implementation of the Analytical Hierarchy Process (AHP) how to link in a hierarchy names order complete()

how comparisons work permutations (a, b): 2 implies a over (?) b by 2 normalizing by values missing values link to paper Optimally completes an incomplete pairwise comparison matrix according to the cyclic coordinates algorithm described in Bozóki, S., Fülöp, J. and Rónyai, L., 'On optimal completion of incomplete pairwise comparison matrices,' Mathematical and Computer Modelling, 52:1–2, 2010, pp. 318-333. (https://doi.org/10.1016/j.mcm.2010.02.047)

precision target differs from node's local/global

Random indices Sets the 'consistency_ratio' property of the Compare object, using random index estimates from Donegan and Dodd's 'A note on Saaty's Random Indexes' in Mathematical and Computer Modelling, 15:10, 1991, pp. 135-137 (doi: 10.1016/0895-7177(91)90098-R) by default (random_index='dd'). If the random index of the object equals 'saaty', uses the estimates from Saaty's Theory And Applications Of The Analytic Network Process, Pittsburgh: RWS Publications, 2005, p. 31.

link to paper when to use saaty

iterations

tolerance

cr override saaty <= 15 dd <= 100

add_children()

report() structure json name weight weights: local, global, target cr ri elements: count, names children: count, names comparisons: input, computed

checks for values over 0 and that all values are numeric

After sitting on the back burner for too long, I'm picking up this code again. I was pleasantly surprised when I was asked to include a license with the repository; I was genuinely shocked to see that it has been forked multiple times!

• My immediate goal is to update everything to Python 3.7.
• Next, I intend to create some test classes, as I remember that it was becoming more and more difficult to make changes and also be sure that I hadn't broken the code elsewhere.
• Third, I'd like to do some refactoring of the code, especially separating out the checks for the matrices into separate sections.
• Fourth, I'd like to revisit how the various objects interrelate and see if it can be done any better. In particular, I'm wondering whether building the hierarchy using a context manager (in the style of PyMC3) would be more user-friendly. Regardless, I still believe the flexibility of the current approach works well when integrating the code into a larger decision-support system, which is my main use case.