Op imizing pa h analysis in mul i-pe spec i e g aphs:
A s udy on he mig a ion om Ne wo kX o g aph- ool
Welbe P. Fe ei a (wpa aizo@posg ad.lncc.b )
An ˆonio T. A. Gomes ([email p o ec ed])∗
LNCC, Pe ´opolis – RJ – B azil
Mapping pa ien ajec o ies is essen ial o unde s anding he unc ioning o he heal h-
ca e sys em. T adi ional models ace limi a ions in ep esen ing mul iple dimensions o ca e
and handling e en epe i ions and empo al in e als. The amewo k in oduced in Rosa’s
doc o al hesis [1] employs Mul iAspec G aphs (MAGs) [2] o ep esen he jou neys
o pa ien s wi h ch onic condi ions, in eg a ing mul iple pe spec i es o ca e. Ini ially imple-
men ed using he Ne wo kX lib a y [3], he amewo k exhibi ed pe o mance limi a ions.
This wo k e ac o s he amewo k using he g aph- ool lib a y [4], which o e s highe
e iciency due o i s C++ implemen a ion and suppo o pa allel compu a ion. This wo k
compa es he wo implemen a ions, sea ching o signi ican pe o mance and scalabili y gains
while p ese ing he ideli y and analy ical capabili ies o he o iginal model.
Pe o mance compa isons we e based on execu ion ime and memo y usage using a eal
pa ien da ase . The g aph c ea ed om his da ase has 473,770 nodes and 1,888,738 edges.
Au oma ed es s in ol ing c i ical ope a ions (g aph building, node il e ing, dimensionali y
educ ion, cen ali ies compu a ion) ensu ed seman ic equi alence be ween implemen a ions.
Ou esul s demons a e he signi ican pe o mance imp o emen achie ed by e ac o -
ing MAG ope a ions om Ne wo kX o g aph- ool. The e ac o ing esul ed in conside able
educ ions in execu ion ime ac oss all es ed unc ions, highligh ing he compu a ional ben-
e i s o using a C++-based back end. No ably, subs an ial educ ions in execu ion ime we e
obse ed in he dimensionali y educ ion (58.9%) and node il e ing (52.9%) s ages.
While g aph- ool is e icien in mos s eps, memo y p o iling e eals ha ope a ions such as
node il e ing and edge econnec ion can lead o a signi ican ly highe memo y usage (24.6%)
han Ne wo kX. These indings unde sco e he impo ance o balancing speed and memo y
demands when dealing wi h la ge-scale, mul i-pe spec i e g aph p ocessing.
[1] Rosa, C. O. C. S. “Complex ne wo ks o model and mine pa ien pa hways.” Doc o al
Thesis, LNCC. 2024. h ps://se a.lncc.b /handle/1/378
[2] Wehmu h, K., Fleu y, E., & Zi iani, A. Mul iAspec G aphs: Algeb aic ep esen a ion
and algo i hms. CoRR, abs/1504.07893. 2015. h p://a xi .o g/abs/1504.07893
[3] Hagbe g, A., Swa , P. J., & Schul , D. A. Ne wo kX – Ne wo k Analysis in Py hon. 2024.
h ps://ne wo kx.o g/
[4] Peixo o, T. P. The g aph- ool Py hon lib a y. 2025. h ps://g aph- ool.skewed.de/
∗The au ho s hank CAPES and he LNCC A i icial In elligence Ins i u e (ins i u o.ia.lncc.b ) o
he inancial suppo .
1
