Ka a, Ahme
A icle
An a i icial in elligence-based o ecas ing o he dynamics o ela i e
p o i a es a a inancial c isis junc u e: A model, a case s udy and
c isis managemen policies
Financial In e ne Qua e ly
P o ided in Coope a ion wi h:
Uni e si y o In o ma ion Technology and Managemen , Rzeszów
Sugges ed Ci a ion: Ka a, Ahme (2025) : An a i icial in elligence-based o ecas ing o he dynamics
o ela i e p o i a es a a inancial c isis junc u e: A model, a case s udy and c isis managemen
policies, Financial In e ne Qua e ly, ISSN 2719-3454, Sciendo, Wa saw, Vol. 21, Iss. 1, pp. 15-26,
h ps://doi.o g/10.2478/ iq -2025-0002
This Ve sion is a ailable a :
h ps://hdl.handle.ne /10419/329893
S anda d-Nu zungsbedingungen:
Die Dokumen e au EconS o dü en zu eigenen wissenscha lichen
Zwecken und zum P i a geb auch gespeiche und kopie we den.
Sie dü en die Dokumen e nich ü ö en liche ode komme zielle
Zwecke e iel äl igen, ö en lich auss ellen, ö en lich zugänglich
machen, e eiben ode ande wei ig nu zen.
So e n die Ve asse die Dokumen e un e Open-Con en -Lizenzen
(insbesonde e CC-Lizenzen) zu Ve ügung ges ell haben soll en,
gel en abweichend on diesen Nu zungsbedingungen die in de do
genann en Lizenz gewäh en Nu zungs ech e.
Te ms o use:
Documen s in EconS o may be sa ed and copied o you pe sonal
and schola ly pu poses.
You a e no o copy documen s o public o comme cial pu poses, o
exhibi he documen s publicly, o make hem publicly a ailable on he
in e ne , o o dis ibu e o o he wise use he documen s in public.
I he documen s ha e been made a ailable unde an Open Con en
Licence (especially C ea i e Commons Licences), you may exe cise
u he usage igh s as speci ied in he indica ed licence.
h ps://c ea i ecommons.o g/licenses/by-nc-nd/3.0/
10.2478/ iq -2025-0002
Abs ac The pu pose o his pape is o (i) demons a e ha he beha io o he ela i e p o i a es a
inancial c isis junc u es in a dual inancial sys em could be di e en han ha o he o he pe i-
ods, (ii) show ha ela i e p o i a es (and hei dynamics) a c isis junc u es could be o ecas -
ed wi h a ela i ely high deg ee o accu acy ia a i icial in elligence algo i hms and (iii) exempli y
he possibili y o c isis-managemen policies ha can smoo hen he ajec o y o he ela i e
p o i a es and acili a e he con ol o possible e a ic luc ua ions a he c isis junc u es in such
sys ems. We employ a se ies o me hodological ools in ol ing (i) s a is ical es s, (ii) a i icial
in elligence algo i hms and (iii) he sys em dynamics simula ion me hod o achie e he h ee
objec i es ou lined in he pa ag aph abo e. The esul s a e o p ac ical signi icance o he inan-
cial policy make s aiming o o mula e and pu in p ac ice e ec i e policies a c isis junc u es.
JEL classi ica ion: G1, C6
Keywo ds: Dual Financial Sys ems, Dynamics o Rela i e P o i Ra es, A i icial-In elligence-Based– Fo ecas ing, C isis Junc u es, C isis Manage-
men
Recei ed: 18.07.2024 Accep ed: 04.11.2024
Ci e his:
Ka a, A. (2025). An a i icial in elligence-based o ecas ing o he dynamics o ela i e p o i a es a a inancial c isis junc u e: A model, a case s udy
and c isis managemen policies. Financial In e ne Qua e ly 21(1), pp. 15-26.
© 2025 Ahme Ka a, published by Sciendo. This wo k is licensed unde he C ea i e Commons A ibu ion-NonComme cial-NoDe i a i es 3.0 Li-
cense.
1 Is anbul Comme ce Uni e si y, Tu key, e-mail: ka a2010@ ica e .edu. , ORCID: h ps://o cid.o g/0000-0002-0162-8137.
Ahme Ka a
An a i icial in elligence-based o ecas ing o he dynamics o ela i e p o i a es a
a inancial c isis junc u e: A model, a case s udy and c isis managemen policies
Financial In e ne Qua e ly 2025, ol. 21 / no. 1
and Çakı (2018). The e ec o he social-poli ical con-
lic s and clashes o ha pe iod on he beha io o he
deposi -holde s a he pa icipa ion banks is ano he
dimension o he sub-sec o al c isis in ques ion, he
de ailed desc ip ion o which is beyond he scope o
his pape . On he o he hand, he e ec o he c isis on
he ela i e p o i a es is a mo e conc e e phenome-
non e lec i e o many o he complexi ies associa ed
wi h he c isis in ques ion. Hence ha is wha we will
ocus on in his pape .
We will analyze his c isis wi h some ask-speci ic
sys ema ic s eps. Following a b ie li e a u e e iew in
he second sec ion o he pape , we unde ake h ee
asks. Fi s , we demons a e he s uc u al di e ence o
he c isis-pe aining beha io o he ela i e p o i a es
om ha o he o he pe iods. Second, in he p esence
o he di icul ies ha he luc ua ion-p one c isis junc-
u es could p esen o he analysis o he opic, we
show ha ela i e p o i a es (and hei dynamics)
could be o ecas ed wi h a ela i ely high deg ee o
accu acy ia a i icial in elligence algo i hms. Finally,
inco po a ing, in o a simula i e amewo k some o he
c isis- ela ed pa e ns and ela ionships iden i ied by
he a i icial in elligence algo i hms, we exempli y he
possibili y o c isis-managemen policies ha can
smoo h he ajec o y o he ela i e p o i a es and
acili a e he con ol o possible e a ic luc ua ions a
he c isis junc u es. The concluding ema ks a e p e-
sen ed in he inal sec ion.
The li e a u e on gene al inancial c ises is o
cou se qui e ich. Alibe e al. (2023) p o ide an in-
o ma i e accoun o many o he c isis- ela ed phe-
nomena such as c ashes, panics, con agion, and possi-
ble policy esponses. The e a e also wo ks explo ing
mo e speci ic issues such as inancial policy e o m
(Hlaing & Kakinaka, 2018), inancial chaos (Dusza,
2017), ins abili y-subop imali y-and chaos-p one luc u-
a ions a c isis junc u es (Ka a, 2023) and global imbal-
ances in ela ion o inancial globaliza ion Liang (2012).
Though he gene al c isis endencies need o be aken
in o accoun by a inance- ela ed wo k such as his, we
should emphasize ha wha we indica e, a gue o , and
analyze in his pape is no he exis ence o a gene al
c isis in he en i e economy du ing ha pe iod bu only
a sub-sec o al c isis wi hin a subse o he banking sys-
em ha was hea ily in luenced by he social-poli ical
clashes in ques ion. Thus, i su ices o p o ide e i-
dence o he c isis in he subsec o in ques ion in he
ele an pe iod. The e en s ha explici ly b oke ou in
2013 and ended in he mili a y coup in 2016 swep
aside 30% o he pa icipa ion banking sys em and in-
luenced he es o he subsys em. This cons i u es
C isis junc u es could pose conside able di icul ies
o he analysis o inancial p ocesses. The easons o
hese di icul ies eside, in pa , in he ex ao dina y
complexi ies ha could be b ough abou by he in e -
ac ions o in ica ely con olu ed mul iple endencies
ha may be p esen a such junc u es. Such endencies
could lead o s uc u al b eaks in pa e ns o inancial
p ocesses and luc ua ions in ma ke s ha may be
p one o ins abili y and e en chaos. In he p esence o
s uc u al b eak-inclusi e possible uns able, e a ic,
and e en chao ic luc ua ions, o ecas ing he ajec o-
ies o inancial p ocesses may p esen signi ican chal-
lenges ha may no be adequa ely handled wi hin he
con ines o he adi ional me hods. I is exac ly he
p esence o such challenges ha calls o he employ-
men o new me hods ha may u n ou o be highly
e ec i e in sol ing some o he p oblems posed by he
challenges in ques ion.
A i icial in elligence algo i hms cons i u e a g oup
o new me hods ha may be wo h employing o he
ask o iden i ying he pa e ns displayed a he compli-
ca ed junc u es in dual inancial sys ems whe e di e -
en ypes o inancial ins i u ions coexis wi hin he
same inancial sys em. Examples o dual inancial sys-
ems could be o mul iple ypes such as sys ems con-
sis ing o s a e and ede al banks o di e en a ie ies
such in he USA o sys ems cha ac e ized by he join
p esence o pa icipa o y banks based on p o i -and-
loss a angemen s and con en ional banks based on
in e es -ea ning a angemen s, which exis in coun ies
such as England, Tu key, Malaysia and Indonesia. We
will explo e he dual sys ems o join pa icipa o y and
con en ional banks and use a case s udy (a pa icula
sub-sec o al c isis junc u e) ela ing o Tu key, he les-
sons o which a e ele an o all coun ies wi h dual
sys ems. The junc u e we will examine in ol ed mul i-
dimensional, di icul - o-analyze clashes be ween he
yea s 2013 and 2016 in Tu key. Du ing he pe iod in
ques ion, Tu key expe ienced a con lic -loaded and
g idlock-p oducing c isis o in ica ely in e linked social,
poli ical and economic dimensions, which esul ed in
an a emp ed coup in July 2016. Needless o say, no all
social-poli ical con lic s and c ises lead o se ious eco-
nomic o inancial consequences o sub-sec o al c isis
p opo ions. Bu his one led o a c isis in he ele an
subse o he banking sys em because he la ges bank
o he pa icipa ion banking sys em (Asya Bank), which
belonged o a sec /communi y and which main ained,
a ha ime, a ound 30% o he pa icipa ion deposi s
in Tu key an in o se ious di icul ies du ing he pe iod
in ques ion and go closed down a e he a emp ed
coup, which was clea e idence o he c isis in ha sub-
pa o he banking sys em in Tu key (Sabah, 2016). Fo
gene al accoun s o he o e all c isis, see: Alkan (2016)
Ahme Ka a
An a i icial in elligence-based o ecas ing o he dynamics o ela i e p o i a es a
a inancial c isis junc u e: A model, a case s udy and c isis managemen policies
Financial In e ne Qua e ly 2025, ol. 21 / no. 1
ma ke p ocesses, which is a highly di icul ask o un-
de ake. How success ul he li e a u e has been in un-
de aking such a ask is an open ques ion. Ne e he-
less, he li e a u e con ains wo ks co e ing chao ic and
s ochas ic issues. Among hese wo ks a e: Zhang
and Wang (2017), Ky sou and Te eza (2002), Dakh-
laoui and Aloui (2013), Gilmo e (1993), Benhabib
(1992), Cohen (1997), as well as Aoki (2001). S ochas ic
and chao ic endencies a e also ela ed o he a ious
dimensions o c isis and policy conside a ions, which
a e examined by wo ks such as Ka a (2023), Ga el and
Pe i -Romec (2017), G osse (2017), He nandez and
Mendoza (2017), Dzhagi yan (2017), Sau (2013), Pu ica
(2015), Be nanke e al. (2019), Minsky (2008) as well as
Ga eo and Molina i (2017). The e a e also po en ially
ui ul sys em dynamics app oaches o he banking
c ises in pa icula and inancial c ises in gene al. Wo ks
in his e ain, such as Kassem and Saleh (2005) and
Schol en (2016), i p ope ly ex ended o cap u e he
in ica e de ails o inancial ela ions, could shed some
ligh on he complica ed eedback s uc u es cha ac e -
izing c isis-p one p ocesses. The e also exis s uc u al
and e olu iona y app oaches desc ibed by Scazzie i
(2018), which can be p oduc i ely used in he analysis
o a ious in ica e de ails o hese p ocesses.
Howe e comp ehensi e he li e a u e’s co e age
may ha e been, c isis junc u es p esen inexhaus ibly
ich p oblems o u he s udy. One such p oblem is
ela ed o he analysis o he ela i e p o i a es (p o i
-in e es a ios), which e lec he dynamics associa ed
wi h he pa icipa ion and con en ional banking sys-
ems, and which we e hea ily in luenced by he clash
and s a egic g idlock expe ienced a he junc u e in
ques ion. In his pape , we analyze he ela i e p o i
a es ia a i icial in elligence algo i hms. The li e a u e
also con ains a i icial in elligence-mo i a ed wo ks on
inancial p ocesses such as U baniko a and S ubno a
(2020), Lin e al. (2017), Ko ol and Fodadis (2022),
To ky and Hassanian (2023) as well as Yang and Chen
(2018), which explo e issues anging om he use o
a i icial neu al ne wo ks in capi al ma ke s o a i icial
in elligence in dynamic en i onmen s. Ou wo k di e s
om hose in he li e a u e in i s unique blend o a i i-
cial in elligence algo i hms and sys em dynamics ha
opens he doo s o a wide a ay o policies we exempli-
y in his pape .
Analogous o he inancial sys ems o some coun-
ies wi h p edominan ly Muslim popula ions, he Tu k-
ish banking sys em includes con en ional as well as
pa icipa ion banks. In con as wi h he usual in e es -
ea ning a angemen s o he con en ional banks, pa -
icipa o y banks’ deposi s/ unds a e based on an
ag eed upon p o i and loss sha ing a angemen
whe e e u ns on deposi s a e de e mined a he end
o he pe iod, depending on he unds’ ealized p o i s
o losses. Whe he he e u ns on deposi s a con en-
su icien e idence o c isis p opo ion in he subsys em
(inclusi e o he pa icipa ion banking) in ques ion in
iew o (and in compa ison wi h) he ac ha an ap-
p oxima ely 25% con ac ion in he banking sys em in
2001 was conside ed one o he mos se ious c ises in
Tu kish banking his o y. O cou se, he c isis in 2001
was an o e all banking c isis while he one we analyze
was a sub-sec o al c isis. Ne e heless, he ela i e con-
ac ion a es we e simila and, as such, i he o me
was a unanimously-ag eed upon c isis, he la e could
and should be conside ed a c isis on he sub-sec o al
scale. Fo a ious indica o s o con ac ion o he bank-
ing sec o in 2001 (see: Çoşkun & Eken, 2015). The pe -
o mance o pa icipa ion banking a e he a emp ed
coup is ano he s o y which is beyond he scope o his
pape . Fo a his o ical assessmen o he place o pa -
icipa ion banking in he Tu kish banking sys em, see:
Us aoğlu (2014) and Va sak (2017). Mo eo e , one de i-
ni ion o banking c isis in he li e a u e e e s o
a si ua ion in which a bank ails o compels he go e n-
men o in e ene (Racickas & Vasiliauskai e, 2012;
IMF, 1998). The c i e ia con ained in his de ini ion a e
clea ly me by he sub-sec o al c isis we examine in his
pape ( he e a e o cou se o he dimensions o inan-
cial c ises which a e explo ed in he li e a u e (see:
Reinha & Rogo , 2014).
The complexi y-d i en di e si y o he beha io al
pa e ns displayed a ha junc u e is ano he indica ion
o he unusual na u e o he pe iod in ques ion. Clea ly,
c isis junc u es could un old and lead o a ich spec-
um o pa e ns and ou comes. Ra ional and seemingly
non- a ional op ions and op imal as well as subop imal
beha io al pa e ns could o en coexis side by side a
such junc u es and help c ea e s able, uns able o e en
chao ic ou comes. Fo ins ance, he al uis ic beha io
o some o he deposi -holde s o Asya bank du ing
ha pe iod was an impo an social phenomenon ha
is wo hy o u he examina ion. Ne e heless, i is no
only he beha io o he indi iduals bu also he ins i u-
ions ha shapes he dynamics o in e ac ions a such
junc u es. The ole o ins i u ions, including ha o he
go e nmen , adds u he complexi ies o he eme -
gence o hose pa e ns and ou comes. The li e a u e
con ains a a ie y o wo ks explo ing in ica e dimen-
sions o hese possibili ies in he usual pe iods as well
as in c isis-pe iods. Fo ins ance, he issues associa ed
wi h in es o beha io and psychology a e co e ed by
a wide ange o wo ks including: Filip e al. (2015), Ho -
mann e al. (2013), Ka io yllas e al. (2017), Klein e al.
(2017), Yang e al. (2017), as well as Gennaioli and
Schlei e (2018). Though pa e ns and ou comes in
ma ke s a e closely connec ed wi h he beha io and
psychology o ma ke pa icipan s, he e a e in ica e
dimensions o chao ic and s ochas ic endencies and
eme gen s uc u es in ma ke s ha may no be com-
ple ely and exclusi ely educed o mic o beha io and
may equi e mic o-mac o-con olu ed ea men o he
Ahme Ka a
An a i icial in elligence-based o ecas ing o he dynamics o ela i e p o i a es a
a inancial c isis junc u e: A model, a case s udy and c isis managemen policies
Financial In e ne Qua e ly 2025, ol. 21 / no. 1
ele an c isis junc u e, could we o ecas i s ajec o y
wi h a high deg ee o accu acy? The answe o his
ques ion is he subjec o he second p oposi ion.
P oposi ion 2: Changes in he ela i e p o i a es
a he c isis junc u e in ques ion could be o ecas ed
wi h a high deg ee o accu acy ia a i icial in elligence
algo i hms.
To p o e he p oposi ion, we will cons uc a ime
se ies o ecas ing model which we will implemen using
he da a mining and machine lea ning p og am WEKA
( o desc ip ions o WEKA- ela ed ools and unc ions,
see: F ank e al., 2022). O he p og ams such as R and
MATLAB could be used as well). In ou model, he
change in ela i e p o i a es will be aken o be he
a ge a iable, which is in luenced by a numbe o a i-
ables such as he change in he closing alues o he
s ock ma ke , he change in he bond a es, he change
in he TL-dolla exchange a es and he change in in la-
ion a es in Tu key, which will be named, along he
lines o he p og am’s e minology, as he “o e lay a i-
ables” ha a e ele an o he o ecas ing o he a ge
a iable. We will add he HP il e - ex ac ed cycle o
he change in he ela i e p o i a es o he lis o he
o e lay a iables.
The main a i icial in elligence/machine lea ning
algo i hm we will choose o he pu pose o o ecas ing
is suppo ec o machine eg ession ( o applica ions
o suppo ec o machines in inance, see: Cao e al.,
2005, Tang & Sheng, 2009 as well as Tang e al., 2009).
Suppo ec o machines could also be used o classi i-
ca ion. The cen al o ganizing gene al p inciple unde ly-
ing he me hod e ol es a ound inding he bes hype -
plane sepa a ing he da a poin s belonging o di e en
classes. This is done by maximizing he ma gin be ween
closes da a poin s in di e en classes. The me hod is
highly lexible and gene ally yields accu a e esul s in
bo h low and high dimensional cases.
We will use a couple o in e ela ed me ics o he
o ecas ing accu acy, he i s o which is “ he no mal-
ized oo mean squa ed e o ” (NRMSE), he de ini ion
o which p e- equi es he concep o he oo mean
squa ed e o (RMSE), which is he squa e oo o he
sum, o e all obse a ions, o he squa es o he di e -
ences be ween he p edic ed and ac ual alues o he
changes in he ela i e p o i a es di ided by he num-
be o obse a ions. NRMSE is de ined as:
The second (in e ela ed) me ic is he deg ee o
o ecas ing accu acy, which is 1-NRMSE.
The high deg ee o o ecas ing accu acy p oduced
by he suppo ec o machine eg ession and he asso-
cia ed ajec o ies o he ela i e p o i a e will be
gi en in Sec ion 3.
en ional and pa icipa o y banks, namely in e es
a es and p o i a es, con e ge o di e ge o e ime is
a poin o discussion in he li e a u e (see: Sa aç
& Ze en, 2015; Sukmana & Ib ahim, 2017). The e is o
cou se a mo e gene al discussion abou he simila i ies
and di e ences be ween he con en ional and pa ici-
pa ion (Islamic) banks (Gassouma e al., 2023). Though
being subjec o simila ma ke o ces may a imes
induce con e gence be ween in e es and p o i a es,
ac o s a ec ing he wo subsys ems o he banking
sys em may di e in some pe iods, esul ing in di e -
gence in he a es in ques ion in some pe iods ( o
a desc ip i e analysis o he a ious aspec s o pa ici-
pa ion banks’ p ac ices, p o i s and some al e na i e
ins umen s, see: Büyükakın and Bilal (2016) and
Gökalp (2014). The e iciency and dis ibu i e implica-
ions o he pa icipa o y a angemen s a e poin ed ou
in Ka a (2001). Among he poin s o pe iods du ing
which such a di e gence could be obse ed a e he sub
-sec o al c isis junc u es, a p ime example o which
occu ed, in Tu key, du ing 2013-2016 – a pe iod cha -
ac e ized by a poli ical and social c isis wi h economic
consequences ha we men ioned in he in oduc ion.
Clea ly he di e gence be ween he p o i and in-
e es a es could be p ope ly cap u ed by he alues o
he ela i e p o i a e, which is de ined as ollows:
whe e p o i and in e es a es a a e he a e ages in
he ele an banking sys ems a .
The da a se o his s udy is a ailable a “Tü kiye
Cumhu iye i Me kez Bankası” (The Cen al Bank o Tu -
key) and “Ka ılım Bankala ı Bi liği” ( he Union o Pa ici-
pa ion Banks in Tu key.) We ha e used he da a o he
o e all pe iod om Janua y 2006 o June 2017 which is
decomposed in o wo subpe iods, namely he sub-
sec o al c isis pe iod, Decembe 2013-Decembe 2016
and he emainde .
In his pape , we will p o e h ee p oposi ions
abou he ela i e p o i a es and he complexi ies
associa ed wi h hem a he c isis junc u e in ques ion.
P oposi ion 1: The dis ibu ion o he ela i e
p o i a es a he c isis junc u e is no he same as he
dis ibu ion o he ela i e p o i a es a o he junc-
u es.
In o he wo ds, we need o show ha , compa ed
o o he pe iods, he ela i e p o i a es beha e di e -
en ly a c isis junc u es. Using he Mann-Whi ney
U es , we p o e he p oposi ion in Sec ion 3. The
Chow es also indica es he p esence o s uc u al
b eaks associa ed wi h he c isis junc u e.
In iew o he di e en dis ibu ion and hence
pa e n he ela i e p o i a es displayed a he ele-
( )
:
p p o i a e a ime
The ela i e p o i a e a in e es a e a ime
=
/
NRMSE RMSE he ange o he changes
in he ela i e p o i a es du ing he c isis pe iod
=
Ahme Ka a
An a i icial in elligence-based o ecas ing o he dynamics o ela i e p o i a es a
a inancial c isis junc u e: A model, a case s udy and c isis managemen policies
Financial In e ne Qua e ly 2025, ol. 21 / no. 1
p o i a e is he s ock a iable, he change o which
will be he low a iable. The changes in he closing
alues, he bond a es, he exchange a es and he in-
la ion a es desc ibed abo e could unc ion as he aux-
ilia y a iables. Depending on he ype o ela ions/
pa e ns e ealed by he a i icial in elligence algo i hm
employed, some algo i hm-selec ed pa icula lagged
alues o he a iables in he model could se e as aux-
ilia y a iables as well. A simula ion diag am ep e-
sen ing he ela ions among he chosen a iables in ou
model will be desc ibed and a pa icula policy-induced
ajec o y o he ela i e p o i a es will be de i ed in
he ollowing sec ion.
Be o e we p oceed wi h he p oo o his p oposi-
ion, we need o i s check whe he he dis ibu ion o
he ela i e p o i a e o he o e all pe iod in ques ion
is no mal. Bo h he Kolmogo o -Smi no and Shapi o-
Wilk es s indica e ha he ela i e p o i a e du ing
he pe iod in ques ion a e no no mally dis ibu ed.
Thus, o p o e he p oposi ion, we need o use a non-
pa ame ic es such as he Mann Whi ney U es . Us-
ing he es in ques ion, we ended up ejec ing he null
hypo hesis s a ing ha he dis ibu ions a e he same
ac oss he pe iods. The ele an SPSS ou pu is as ol-
lows (Table 1).
The success in o ecas ing, which is impo an in
i sel , also pa es he way o success ul c isis manage-
men -possibili ies/policies. The ela ionships and
pa e ns e ealed o iden i ied du ing he p ocess o
o ecas ing could be used o con ol he ajec o ies o
he inancial p ocesses, which is he subjec o he hi d
p oposi ion.
P oposi ion 3. The e exis ax policies ha can
smoo hen he ela i e p o i ajec o y a c isis junc-
u es.
The me hod we will use o p o e his p oposi ion is
one o a a i icial in elligence-in eg a ed sys em dy-
namics simula ion amewo k whe e we will inco po-
a e he esul s o a pa icula ly simple, easy- o-
in e p e a i icial in elligence algo i hm, such as linea
eg ession, in o a simula i e model and unde ake poli-
cy simula ions.
The sys em dynamics simula ion me hod could be
simply desc ibed in e ms o ou di e en compo-
nen s, namely he “s ock a iables” which ep esen
he accumula ed alues o some key a iables, he
“ low a iables” which ep esen he changes in he
s ock a iables, he “auxilia y a iables” which in lu-
ence (o a e in luenced by) he o he a iables in he
sys em and he “ eedback ela ions” o connec ions
which desc ibe he ela ions o chains o causal de-
pendencies among he a iables in sys em. A p ope
o mula ion o a sys em dynamics model would equi e
he speci ica ion o some o all o hese componen s. In
he illus a i e model we will cons uc , he ela i e
Table 1a: The Mann Whi ney U es esul s
Asymp o ic signi icances a e displayed. The signi icance le el is 0.05
Sou ce: Au ho 's own wo k (p oduced wi h SPSS).
he c isis pe iod, leading o he ejec ion o he no-
b eak hypo hesis.
Al e na i ely, we employed he Chow es which
also indica es ha he e is a b eak a he beginning o
Chow B eakpoin Tes : 2013M12
Null Hypo hesis: No b eaks a speci ied b eakpoin s
Va ying eg esso s: All equa ion a iables
Equa ion Sample: 2006M01 2016M01
F-s a is ic 88.20419 P ob. F (1,119) 0.0000
Log likelihood a io 67.10433 P ob. Chi-Squa e (1) 0.0000
Wald S a is ic 88.20419 P ob. Chi-Squa e (1) 0.0000
Table 1b: The Chow Tes esul s
Sou ce: Au ho 's own wo k (p oduced wi h SPSS).
Null Hypo hesis Tes Sig. Decision
The dis ibu ion o
ela i e p o i a e is he
same ac oss ca ego ies o
wo pe iods
Independen - samples
Mann Whi ney U Tes 0.000 Rejec ed he null
hypo hesis
Ahme Ka a
An a i icial in elligence-based o ecas ing o he dynamics o ela i e p o i a es a
a inancial c isis junc u e: A model, a case s udy and c isis managemen policies
Financial In e ne Qua e ly 2025, ol. 21 / no. 1
change in he ela i e p o i a es. The esul s yield
a NRMSE alue o 0.08 o he es ing da a and hence
a deg ee o o ecas ing accu acy o 0.92 (92%), which is
ai ly high, indica ing success ul o ecas ing. The ajec-
o ies associa ed wi h his o ecas ing a e gi en in Fig-
u e 1 and Figu e 2, whe e “ ela i e p o i a e” indi-
ca es he change in he ela i e p o i a es.
We spli /decomposed he da a in o wo subse s
and used 70% o aining and 30% o es ing. Wi h he
p og am WEKA, we un he a ibu e-selec ed suppo
ec o machine eg ession algo i hm, which cap u es
he unde lying dynamics and enables us o o ecas he
Figu e 1: The o ecas ing o he change in ela i e p o i a es wi h he es ing da a (p oduced wi h WEKA)
Sou ce: Au ho ’s own wo k.
The nume ical esul s indica e ha u u e alues o
he ela i e p o i a es could be o ecas ed wi h a 92%
accu acy a he junc u e unde s udy.
As displayed in Figu e 1, he chosen a i icial in elli-
gence/machine lea ning algo i hm ( he suppo ec o
machine eg ession) yields accu a e o ecas s o he
change in ela i e p o i a e wi h he es ing da a.
Figu e 2: The o ecas ing o he change in ela i e p o i a es wi h he aining da a (p oduced wi h WEKA)
Sou ce: Au ho ’s own wo k.
M5P algo i hm accu a ely o ecas s, as shown in
Figu e 2, he change in he ela i e p o i a e and he
change in he s ock ma ke e u n.
As an ex ension, Figu e 2a displays he ajec o ies
o he change in he ela i e p o i a e and he change
in he s ock ma ke e u n. The ajec o ies a e ob-
ained ia M5P algo i hm.
Ahme Ka a
An a i icial in elligence-based o ecas ing o he dynamics o ela i e p o i a es a
a inancial c isis junc u e: A model, a case s udy and c isis managemen policies
Financial In e ne Qua e ly 2025, ol. 21 / no. 1
Fo he sake o con enience, le us choose he line-
a eg ession algo i hm wi h he aining da a and in-
co po a e he esul s i gene a es in o a sys em dynam-
ics simula ion model. We ha e used he se o o e lay
a iables including he changes in he closing alues,
he bond a es, he exchange a es and he in la ion
a es o es ima e and o ecas he a ge a iable.
The lag- ela ed e ms symbolize he lagged alues
o he ele an a iable o he indica ed pe iods. The
equa ion indica es ha , in addi ion o he “dbond a e”
and “dexchange a e”, d ela i e p o i alues which a e
lagged o 1, 5, 6, 7 and 9 pe iods in luence he a ge
a iable. Simila ly, 4 h, 6 h, 7 h and 9 h lagged alues (o
he ela i e p o i a es) imes an a i icial ime index-
a iable ha e some in luence on he dependen a ia-
ble as well. We can cons uc a simula ion model inco -
po a ing hese in luences. To demons a e ha , in ad-
di ion o hese in luences, he policy we will p opose
wo ks e en in he p esence o shocks cha ac e is ic o
c isis junc u es, we can in oduce o he model some
addi ional complexi y by adding some s ochas ic e ms
o he “dbond a e” and “dexchange a e” a iables.
These ela ions a e inco po a ed in o he ollowing
sys em dynamics simula ion diag am (Figu e 3) which
we ha e d awn using he p og am VENSIM. We can
ake al e na i e ou es o he analy ical kind exempli-
ied in Ka a (2007, 2013) as well as Ka a and Osman
(2006). Howe e , o su icien ly complex s ochas ic
e ms and eedback ela ions, he e may no exis ana-
ly ical solu ions.
To p o e he p oposi ion sugges ing policy-induced
changes a he c isis junc u e, we need o ind a policy
ha leads o a measu able smoo hing o he ajec o y
o he change in ela i e p o i a e. The algo i hm-
p oduced es ima ion ob ained wi h WEKA is as ollows
(To be compa ible wi h he p og am’s allowed symboli-
za ion, we will omi he spaces be ween he wo ds in
a iable names). The es ima ed equa ion cap u es he
dynamics unde lying he change in ela i e p o i a es:
Figu e 2a: The o ecas ing o he changes in he ela i e p o i a e and s ock ma ke e u n wi h aining da a
(p oduced wi h WEKA)
Sou ce: Au ho ’s own wo k.
0.352 0.0244
0.1893 0.002
0.2972 _ 1 0.2825
_ 5 0.7103
_6
0.288
D ela i ep o i a e dbond a e
dexchange a e A i icialTimeIndex
Lag d ela i ep o i a e
Lag d ela i ep o i a e
Lag d ela i ep o i a e
= − + − +
− + +
− + −
− + −
−+
−6 _ 7
0.545 _ 9
0.0353 _ 4
0.0436 _ 6
0.0327 _
Lag d ela i ep o i a e
Lag d ela i ep o i a e
A i icialTimeIndex Lag d ela i ep o i
A i icialTimeIndex Lag d ela i ep o i a e
A i icialTimeIndex Lag d ela
− +
− − +
− − +
− +
7
0.0284 _ 9
i ep o i a e
A i icialTimeIndex Lag d ela i ep o i a e
−+
−
Ahme Ka a
An a i icial in elligence-based o ecas ing o he dynamics o ela i e p o i a es a
a inancial c isis junc u e: A model, a case s udy and c isis managemen policies
Financial In e ne Qua e ly 2025, ol. 21 / no. 1
We will now in oduce/design a ax ha o ces he
change in he ela i e p o i a e downwa ds by 2.5%
when he ela i e p o i a e is g ea e han 1 and up-
wa ds when i is lowe han 1. This ax leads o he a-
jec o y gi en in Figu e 4. The ajec o y wi hou any ax
is displayed in Figu e 5.
Fo he sake o illus a ion, we will assume ha he
“shock-in luenced dbond a e” = ”dbond a e”-0.2u,
whe e u is a uni o mly dis ibu ed andom a iable
aking alues be ween -5% and +5% o he ela i e
p o i a e. Simila ly, he “shock-in luenced dex-
change a e” = “dexchange a e”+u (Al e na i e nume i-
cal alues could be aken o exempli ica ion pu poses).
Figu e 3: The simula ion diag am
Sou ce: Au ho 's own wo k (p oduced wi h VENSIM).
Figu e 4: The ela i e p o i ajec o y wi h he ax (p oduced wi h VENSIM)
Sou ce: Au ho 's own wo k (p oduced wi h VENSIM).
