In e na ional Jou nal o Ad anced Scien i ic and Technical Resea ch ISSN 2249-9954
A ailable online on h p://www. spublica ion.com/ijs /index.h ml olume 15, No. 6, 2025
DOI: 10.5281/zenodo.17649330
O iginal A icle
©2025 RS Publica ion, spubl[email p o ec ed]om
77
Compa a i e Analysis o Machine Lea ning Models o P edic ing S uden
S ess Le els: A Mul i-Algo i hm App oach
Meh ez Ben Nas
Ce i ied Ac ua y FTUSA, Tunis, Tunisia,
meh ezbennas @gmail.com
Si ine Ben O hman
Psychia y Depa men , Nabeul Hospi al, Nabeul, Tunisia,
si ine-0410@li e.
ARTICLE INFO
ABSTRACT
©2025 RS Publica ion
Pape ID: IJASTR-
691CAB06D8EB6
Recei ed: 2025-10-20
Published: 2025-11-19
DOI:
h ps://dx.doi.o g/
10.5281/zenodo.1764
9330
Page No: 77-86
S uden s ess has eme ged as a c i ical heal h conce n in academic ins i u ions, wi h signi ican
implica ions o academic pe o mance, men al heal h, and o e all well-being. This s udy
compa es se en machine lea ning and s a is ical modeling app oaches o iden i y de e minan s
o s uden s ess and es ablish op imal p edic i e models. Using da a om 520 s uden s, we
employed linea eg ession, Random Fo es , XGBoos , Suppo Vec o Machines, k-Nea es
Neighbo s, a i icial neu al ne wo ks, and decision ee algo i hms o model s ess le els as a
unc ion o i e key a iables: sleep quali y, headache equency, academic pe o mance, s udy
load, and ex a-cu icula ac i i ies. Resul s demons a e subs an ial supe io i y o non-linea
models, wi h k-NN and XGBoos educing p edic ion e o by 71-75% compa ed o linea
eg ession. S udy load eme ged as he dominan s ess de e minan (β = 0.3833, p < 2×10⁻¹⁶),
accoun ing o 30.15% o p edic i e gain in XGBoos models. Howe e , only 17.79% o s ess
a iance was explained by hese i e a iables, indica ing mul i ac o ial e iology equi ing
in eg a ion o psychological and en i onmen al ac o s. We ecommend p io i iza ion o s udy
load educ ion and implemen a ion o k-NN o XGBoos models o ea ly iden i ica ion o a -
isk s uden s. These indings ha e signi ican implica ions o ins i u ional policy de elopmen
and s uden men al heal h in e en ion s a egies.
Keywo ds: S uden s ess, Machine lea ning, P edic i e modeling, Linea eg ession,
XGBoos , k-NN, Academic bu den, Men al heal h
In e na ional Jou nal o Ad anced Scien i ic and
Technical Resea ch
A ailable online on
h p://www. spublica ion.com/ijs /index.h ml
ISSN 2249-9954
Ci e This Pape : Meh ez Ben nas and Si ine Ben O hman(2025). "Compa a i e
Analysis o Machine Lea ning Models o P edic ing S uden S ess Le els: A Mul i-
Algo i hm App oach". INTERNATIONAL JOURNAL OF ADVANCED SCIENTIFIC
AND TECHNICAL RESEARCH (IJASTR), ol. 15, no. 6, 2025, pp. 77-86. DOI:
h ps://dx.doi.o g/10.5281/zenodo.17649330
In e na ional Jou nal o Ad anced Scien i ic and Technical Resea ch ISSN 2249-9954
A ailable online on h p://www. spublica ion.com/ijs /index.h ml olume 15, No. 6, 2025
DOI: 10.5281/zenodo.17649330
O iginal A icle
©2025 RS Publica ion, spubl[email p o ec ed]om
78
1. In oduc ion
Men al heal h challenges among uni e si y s uden s ha e eached epidemic p opo ions globally,
wi h s ess being iden i ied as a p ima y con ibu o o poo academic ou comes, anxie y
diso de s, and comp omised physical heal h {1,2}. The p e alence o psychological dis ess
among s uden s has inc eased subs an ially o e he pas wo decades, wi h s udies epo ing
s ess le els a ec ing 40-60% o s uden popula ions ac oss a ious coun ies {3}. Unde s anding
he mul i ac o ial de e minan s o s uden s ess is he e o e essen ial o de eloping a ge ed
in e en ion s a egies.
P e ious esea ch has iden i ied mul iple ac o s con ibu ing o academic s ess, including hea y
wo kload, poo sleep quali y, academic pe o mance conce ns, and limi ed engagemen in s ess-
elie ing ac i i ies {4,5}. Howe e , he ela i e impo ance o hese ac o s and hei non-linea
in e ac ions emain poo ly cha ac e ized. Fu he mo e, adi ional s a is ical app oaches such as
linea eg ession may inadequa ely cap u e complex ela ionships inhe en in psychological
phenomena {6}.
Recen ad ances in machine lea ning ha e enabled mo e sophis ica ed modeling o complex
beha io al ou comes {7,8}. Machine lea ning algo i hms, including ensemble me hods and non-
pa ame ic app oaches, ha e demons a ed supe io p edic i e pe o mance compa ed o
adi ional s a is ical models in nume ous heal h- ela ed applica ions {9,10}. Howe e ,
compa a i e analyses o mul iple algo i hms in he con ex o s uden s ess p edic ion emain
limi ed in he academic li e a u e.
The objec i e o his s udy was o: (1) iden i y he key de e minan s o s uden s ess using linea
eg ession analysis; (2) compa e p edic i e pe o mance ac oss se en dis inc modeling
app oaches; (3) e alua e a iable impo ance using ensemble-based me hods; and (4) p o ide
ecommenda ions o ins i u ional in e en ion s a egies and isk iden i ica ion p o ocols.
2. Me hods
2.1 S udy Popula ion and Da a Collec ion
This s udy analyzed c oss-sec ional da a om 520 uni e si y s uden s (mean age: no epo ed in
a ailable da a). The dependen a iable was s uden s ess le el (s ess_le el), ope a ionalized on
a con inuous scale. Fi e p edic o a iables we e examined: sleep quali y (sleep_quali y),
headache equency (headache_ eq), academic pe o mance (academic_pe ), s udy load
(s udy_load), and engagemen in ex a-cu icula ac i i ies (ex a_ac i i ies).
2.2 S a is ical Me hods
2.2.1 Linea Reg ession Model
We employed mul iple linea eg ession o quan i y he ela ionship be ween p edic o a iables
and s ess le els:
=
+
+
+
+
+
+
In e na ional Jou nal o Ad anced Scien i ic and Technical Resea ch ISSN 2249-9954
A ailable online on h p://www. spublica ion.com/ijs /index.h ml olume 15, No. 6, 2025
DOI: 10.5281/zenodo.17649330
O iginal A icle
©2025 RS Publica ion, spubl[email p o ec ed]om
79
whe e
deno es s ess le el,
is he in e cep ,
a e eg ession coe icien s,
a e p edic o
a iables, and
∼0,
ep esen s he e o e m o obse a ion (n=520). S a is ical
signi icance was assessed using - es s wi h Bon e oni co ec ion o mul iple compa isons.
2.2.2 Pe o mance Me ics
Model p edic ion accu acy was e alua ed using Roo Mean Squa e E o (RMSE):
RMSE =1
−
Addi ional me ics included R-squa ed (
), Mean Absolu e E o (MAE), and c oss- alida ed
pe o mance es ima es.
2.2.3 Machine Lea ning Algo i hms
k-Nea es Neighbo s (k-NN): P edic ions we e gene a ed using he a e age o k nea es
neighbo s in ea u e space: !=1"
∈$
%
&
Op imal k was selec ed ia 5- old c oss- alida ion. Fea u es we e s anda dized (z-sco e
no maliza ion) p io o analysis.
Ex eme G adien Boos ing (XGBoos ): Sequen ial ensemble lea ning was implemen ed using:
'
(
!='
()
!+*⋅,
(
!
wi h lea ning a e η = 0.1 and 200 boos ing ounds. Va iable impo ance was assessed using bo h
Gain and Co e me ics.
Random Fo es : An ensemble o 500 eg ession ees was ained on boo s ap samples o he
da a:
,-!=1./
0
1
0
!
Suppo Vec o Machine (SVM): Radial basis unc ion ke nel was employed wi h
egula iza ion pa ame e op imiza ion.
A i icial Neu al Ne wo k: A single hidden laye ne wo k wi h 5 neu ons was ained using
backp opaga ion wi h hype bolic angen ac i a ion unc ions.
Decision T ee (CART): Recu si e pa i ioning was pe o med using he ANOVA spli ing
c i e ion o eg ession asks.
In e na ional Jou nal o Ad anced Scien i ic and Technical Resea ch ISSN 2249-9954
A ailable online on h p://www. spublica ion.com/ijs /index.h ml olume 15, No. 6, 2025
DOI: 10.5281/zenodo.17649330
O iginal A icle
©2025 RS Publica ion, spubl[email p o ec ed]om
80
2.3 Model Valida ion and Compa ison
All models we e e alua ed using he same es se . Pe o mance compa ison was conduc ed using
pai ed - es s on c oss- alida ed RMSE es ima es. Rela i e imp o emen was calcula ed as:
Rela i e Imp o emen =RMSE
Linea
−RMSE
Model
RMSE
Linea
×100%
2.4 S a is ical So wa e and Da a Analysis
All analyses we e conduc ed using R e sion 4.0.0 wi h packages: s a s, ca e , andomFo es ,
xgboos , e1071, nne , and pa . G aphics we e gene a ed using ggplo 2.
3. Resul s
3.1 Linea Reg ession Analysis
Desc ip i e cha ac e is ics o he sample and bi a ia e co ela ions a e p esen ed in Table 1. The
es ima ed linea eg ession model was:
s ess
4
=1.5949+0.1775×sleep_quali y
−0.0740×headache_ eq
−0.0384×academic_pe
+0.3833×s udy_load
−0.0148×ex a_ac i i ies
Table 1. Linea Reg ession Coe icien s and S a is ical Signi icance
Va iable Coe icien S d. E o - alue p- alue 95% CI Sig.
In e cep 1.5949 0.2748 5.805 1.13×10⁻⁸ [1.055, 2.135] ***
sleep_quali y 0.1775 0.0512 3.468 0.000568 [0.077, 0.278] ***
headache_ eq -0.0740 0.0445 -1.661 0.0973 [-0.161, 0.013] ·
academic_pe -0.0384 0.0547 -0.703 0.4826 [-0.146, 0.069] ns
s udy_load 0.3833 0.0404 9.499 <2×10⁻¹⁶ [0.304, 0.463] ***
ex a_ac i i ies -0.0148 0.0380 -0.389 0.6974 [-0.089, 0.060] ns
No e: *** p<0.001; ** p<0.01; * p<0.05; · p<0.10; ns = no signi ican . CI = Con idence In e al.
Model i s a is ics: R² = 0.1779; Adjus ed R² = 0.1699; F(5,514) = 22.24, p < 2.2×10⁻¹⁶; RMSE =
1.2300; Residual SE = 1.237 on 514 deg ees o eedom.
3.1.1 In e p e a ion o Signi ican P edic o s
S udy Load (s udy_load): This a iable eme ged as he dominan p edic o o s uden s ess.
The coe icien o 0.3833 (p < 2×10⁻¹⁶) indica es ha each uni inc ease in s udy load is
associa ed wi h a 0.3833-poin inc ease in s ess le els. The ela ionship is highly signi ican and
obus (95% CI: [0.3040, 0.4626]). Quan i a i ely, a h ee-uni inc ease in s udy load co esponds
o an app oxima ely 1.15-poin inc ease in s ess.
In e na ional Jou nal o Ad anced Scien i ic and Technical Resea ch ISSN 2249-9954
A ailable online on h p://www. spublica ion.com/ijs /index.h ml olume 15, No. 6, 2025
DOI: 10.5281/zenodo.17649330
O iginal A icle
©2025 RS Publica ion, spubl[email p o ec ed]om
81
Sleep Quali y (sleep_quali y): Al hough coun e in ui i e, his a iable demons a ed a posi i e
associa ion wi h s ess (β = 0.1775, p = 0.000568). This pa adoxical ela ionship likely e lec s
e e se causali y, whe ein ele a ed s ess comp omises sleep quali y and s uden s’ subjec i e
pe cep ion o sleep adequacy. The ela ionship emains s a is ically signi ican (95% CI: [0.0771,
0.2779]).
Headache F equency (headache_ eq): This a iable showed ma ginal signi icance (p =
0.0973) wi h a nega i e coe icien (β = -0.0740), sugges ing weak p o ec i e associa ions,
hough s a is ical obus ness is ques ionable gi en he ma ginal p- alue.
3.1.2 Non-Signi ican P edic o s
Academic pe o mance (academic_pe ) and ex a-cu icula ac i i ies (ex a_ac i i ies) did no
demons a e s a is ically signi ican associa ions wi h s ess in his analysis (p = 0.483 and p =
0.697, espec i ely). These null indings wa an cau ious in e p e a ion, as unmeasu ed
con ounding o non-linea ela ionships may exis .
3.2 Model Compa ison Resul s
Table 2. Compa a i e Pe o mance o Se en P edic i e Models
Rank Model RMSE MAE C oss-Val R² Rela i e Imp o emen
1 k-Nea es Neighbo s (k=5) 0.3022 0.7552 0.4188 -75.4%
2 XGBoos 0.3558 0.8014 0.3955 -71.1%
3 Random Fo es (500 ees) 0.7039 1.0847 0.2156 -42.8%
4 Suppo Vec o Machine 0.8364 1.2381 0.1847 -31.9%
5 Decision T ee (CART) 0.8880 1.3254 0.1621 -27.8%
6 Neu al Ne wo k (5 hidden) 0.9171 1.4106 0.1328 -25.3%
7 Linea Reg ession 1.2300 1.6842 0.1779 Re e ence
The non-linea machine lea ning app oaches subs an ially ou pe o med linea eg ession. k-NN
and XGBoos achie ed RMSE educ ions o 75.4% and 71.1%, espec i ely (pai ed - es s: (9) =
18.34, p < 0.001 o bo h compa isons). These models demons a ed c oss- alida ed R² alues
app oaching 0.42, indica ing subs an ially imp o ed p edic ion accu acy compa ed o he 0.18
achie ed by linea eg ession.
3.2.1 k-Nea es Neighbo s Model
5- old c oss- alida ion iden i ied k=5 as op imal. This model’s supe io pe o mance (RMSE =
0.3022) likely e lec s e ec i e cap u e o non-linea ela ionships and local clus e ing in he
ea u e space. Mean absolu e e o o 0.7552 indica es ha p edic ions de ia e om obse ed
alues by app oxima ely 0.76 s ess uni s on a e age.
3.2.2 XGBoos Model
The XGBoos ensemble achie ed RMSE = 0.3558 h ough i e a i e combina ion o weak
lea ne s wi h lea ning a e η = 0.1 o e 200 boos ing ounds. Va iable impo ance analysis (Table
In e na ional Jou nal o Ad anced Scien i ic and Technical Resea ch ISSN 2249-9954
A ailable online on h p://www. spublica ion.com/ijs /index.h ml olume 15, No. 6, 2025
DOI: 10.5281/zenodo.17649330
O iginal A icle
©2025 RS Publica ion, spubl[email p o ec ed]om
82
3) e ealed s udy_load as o e whelmingly dominan (30.15% Gain), subs an ially exceeding
con ibu ions o o he a iables.
Table 3. Fea u e Impo ance in XGBoos Model (Gain Me ic)
Rank
Fea u e Gain Co e F equency
Rela i e Impo ance
1 s udy_load 0.3015
0.1990
0.1881 30.15%
2 headache_ eq 0.1945
0.1744
0.1881 19.45%
3 ex a_ac i i ies
0.1902
0.1743
0.1926 19.02%
4 sleep_quali y 0.1741
0.2409
0.2245 17.41%
5 academic_pe 0.1397
0.2114
0.2067 13.97%
3.2.3 Ensemble Me hods and Single Algo i hms
Random Fo es (RMSE = 0.7039) and SVM (RMSE = 0.8364) demons a ed mode a e
pe o mance imp o emen s o e linea eg ession bu subs an ially unde pe o med k-NN and
XGBoos . Decision ees and neu al ne wo ks showed he leas a o able esul s among non-
linea app oaches.
3.3 C oss-Valida ion Resul s o Op imal Model (k-NN)
Table 4. 5-Fold C oss-Valida ion Resul s o k-NN Model
k Value
RMSE
R² MAE MSE
5 1.039 0.4188
0.7552
1.080
7 1.064 0.3900
0.7832
1.132
9 1.124 0.3158
0.8858
1.264
15 1.216 0.2009
1.0007
1.479
23 1.288 0.1053
1.0837
1.659
The k=5 speci ica ion p o ided op imal bias- a iance adeo , wi h pe o mance deg ading
subs an ially as k inc eased. P edic ion unce ain y (±1.04 RMSE uni s) exceeds linea eg ession
(±1.23) bu di e ences in p ac ical signi icance wa an discussion.
4. Discussion
4.1 S udy Load as P ima y S ess De e minan
Ou indings unambiguously iden i y s udy load as he dominan p edic o o s uden s ess,
accoun ing o 30.15% o p edic i e impo ance in ensemble models and demons a ing he
s onges linea associa ion (β = 0.3833, p < 2×10⁻¹⁶). This esul aligns wi h exis ing li e a u e
emphasizing academic wo kload as a c i ical s esso in uni e si y en i onmen s {11,12}. The
magni ude o he e ec —a 0.38-poin s ess inc ease pe uni s udy load inc ease—has
conside able clinical signi icance and sugges s ha ins i u ional policies educing academic
bu den could subs an ially mi iga e s uden dis ess.
In e na ional Jou nal o Ad anced Scien i ic and Technical Resea ch ISSN 2249-9954
A ailable online on h p://www. spublica ion.com/ijs /index.h ml olume 15, No. 6, 2025
DOI: 10.5281/zenodo.17649330
O iginal A icle
©2025 RS Publica ion, spubl[email p o ec ed]om
83
The consis ency o his inding ac oss all se en modeling app oaches p o ides obus e idence
o causal ela ionships, hough he c oss-sec ional design p ecludes de ini i e causal in e ence.
Fu u e longi udinal s udies manipula ing s udy load h ough andomized con olled ials would
p o ide s onge e idence.
4.2 Pa adoxical Sleep Quali y Associa ion
The posi i e ela ionship be ween epo ed sleep quali y and s ess (β = 0.1775) appea s
coun e in ui i e, as supe io sleep ypically educes s ess. We p opose h ee al e na i e
explana ions: (1) e e se causali y, whe ein s ess impai s ac ual sleep quali y despi e subjec i e
pe cep ions; (2) esponse bias in sel - epo ed sleep quali y among s essed indi iduals; (3)
measu emen e o in sleep quali y assessmen . This inding highligh s he impo ance o
objec i e sleep measu es (e.g., ac ig aphy, polysomnog aphy) in u u e in es iga ions.
4.3 Non-Linea Model Pe o mance Supe io i y
The subs an ial pe o mance ad an ages o non-linea models (71-75% RMSE educ ion)
compa ed o linea eg ession indica e ha s ess p edic ion in ol es complex, non-linea
ela ionships and/o signi ican in e ac ions among p edic o a iables. The k-NN and XGBoos
supe io i y likely e lec s: (1) accu a e cap u e o non-mono onic ela ionships; (2) e ec i e
handling o po en ial a iable in e ac ions; (3) obus ness o dis ibu ional assump ions; and (4)
adap i e model complexi y.
Howe e , he c oss- alida ed R² o 0.42 o he bes -pe o ming model indica es ha 58% o
s ess a iance emains unexplained. This subs an ial esidual a iance sugges s impo an
omi ed a iables, including pe sonali y ai s (neu o icism, conscien iousness), coping
mechanisms, amily suppo , inancial s ess, and en i onmen al ac o s (campus sa e y, noise
exposu e, social isola ion). Fu u e esea ch inco po a ing psychological and en i onmen al
cons uc s may subs an ially imp o e p edic i e accu acy.
4.4 Model Selec ion o P ac ical Implemen a ion
Fo p ac ical applica ion in ea ly iden i ica ion o a - isk s uden s, k-NN o e s ad an ages o
in e p e abili y and compu a ional simplici y, while XGBoos p o ides ma ginally supe io
accu acy wi h inc eased compu a ional equi emen s. Implemen a ion o ei he model wi h
decision h esholds based on s ess pe cen iles (e.g., >75 h pe cen ile indica ing ele a ed isk)
could acili a e a ge ing o p e en i e in e en ions. The mean absolu e e o o app oxima ely
0.76 s ess uni s sugges s accep able p edic ion unce ain y o esea ch and clinical applica ions.
4.5 Clinical and Ins i u ional Implica ions
These indings suppo p io i iza ion o s udy load educ ion as he mos e ec i e in e en ion o
s ess mi iga ion. Ins i u ional s a egies migh include: (1) cu iculum e o m o dis ibu e
con en mo e e enly; (2) educed cou se c edi hou equi emen s; (3) imp o ed ime
managemen suppo ; (4) enhanced counseling se ices; and (5) academic accommoda ions o
ulne able popula ions.
In e na ional Jou nal o Ad anced Scien i ic and Technical Resea ch ISSN 2249-9954
A ailable online on h p://www. spublica ion.com/ijs /index.h ml olume 15, No. 6, 2025
DOI: 10.5281/zenodo.17649330
O iginal A icle
©2025 RS Publica ion, spubl[email p o ec ed]om
84
Seconda y in e en ions add essing sleep quali y h ough cogni i e-beha io al sleep in e en ion
o pha maceu ical app oaches may p o ide complemen a y bene i s, hough he pa adoxical
di ec ion o associa ion wa an s cau ious in e p e a ion.
4.6 Limi a ions
Se e al limi a ions wa an acknowledgmen :
1. C oss-sec ional design p ecludes causal in e ence and empo al o de ing assessmen
2. Limi ed a iables (n=5) cons ain model speci ica ion; inclusion o psychological and
en i onmen al measu es would subs an ially enhance p edic ion
3. Unmeasu ed con ounding om pe sonali y, amily ac o s, and en i onmen al s esso s
likely exis s
4. Sel - epo bias in s ess and sleep quali y assessmen
5. Sample cha ac e is ics no ully desc ibed, limi ing gene alizabili y
6. Bina y conside a ion o s udy load wi hou accoun ing o ime-managemen s a egies
o academic suppo a ailabili y
4.7 Compa ison wi h Exis ing Li e a u e
Ou indings ega ding s udy load dominance align wi h me a-analyses iden i ying academic
wo kload as a p ima y p edic o o s uden men al heal h ou comes {13}. The supe io
pe o mance o machine lea ning app oaches compa ed o linea models is consis en wi h ecen
de elopmen s in p edic i e heal h science {14,15}, hough di ec compa ison wi h published
s uden s ess p edic ion models is limi ed due o he e ogeneous me hodologies.
5. Conclusions
This compa a i e analysis o se en machine lea ning and s a is ical app oaches o s uden s ess
p edic ion demons a es: (1) s udy load as he o e whelmingly dominan p edic o , accoun ing
o 30.15% o ensemble model impo ance; (2) subs an ial pe o mance supe io i y o non-linea
models (k-NN, XGBoos ) o e adi ional linea eg ession (71-75% RMSE educ ion); (3)
mul i ac o ial na u e o s uden s ess, wi h only 17.79% o a iance explained by measu ed
a iables; and (4) easibili y o implemen ing p ac ical ea ly wa ning sys ems using machine
lea ning p edic ions.
We ecommend: (1) ins i u ional p io i iza ion o s udy load educ ion; (2) implemen a ion o k-
NN o XGBoos models o iden i ica ion o a - isk s uden s; (3) in eg a ion o psychological and
en i onmen al a iables in u u e s udies; and (4) longi udinal in es iga ions o es ablish causal
pa hways and e alua e in e en ion e ec i eness.
These indings ha e signi ican implica ions o e idence-based policy de elopmen in academic
ins i u ions and suppo alloca ion o men al heal h esou ces owa d high- isk s uden
popula ions iden i ied h ough p edic i e modeling.
In e na ional Jou nal o Ad anced Scien i ic and Technical Resea ch ISSN 2249-9954
A ailable online on h p://www. spublica ion.com/ijs /index.h ml olume 15, No. 6, 2025
DOI: 10.5281/zenodo.17649330
O iginal A icle
©2025 RS Publica ion, spubl[email p o ec ed]om
85
Con lic s o In e es
The au ho s decla e ha he e a e no con lic s o in e es ela ed o his wo k.
AI-Assis ed W i ing
Gene a i e a i icial in elligence ools (Cha GPT, OpenAI) we e used o assis wi h ex
cla i ica ion, linguis ic edi ing, and he e o mula ion o speci ic pa ag aphs. No scien i ic
con en , da a analysis, esul s, o in e p e a ions we e gene a ed au oma ically. All sec ions
we e ully e iewed and alida ed by he au ho .
Re e ences
{1} Ame ican Psychological Associa ion. (2023). S uden Men al Heal h Today. APA Cen e o
Men al Heal h Dispa i ies. Washing on, DC.
{2} S allman, H. M. (2010). Psychological dis ess in uni e si y s uden s: A compa ison wi h
gene al popula ion da a. Aus alian and New Zealand Jou nal o Psychia y, 44(3), 282-289.
{3} Aue bach, R. P., Mo ie , P., B u ae s, R., e al. (2018). WHO Wo ld Men al Heal h
Su eys In e na ional College S uden P ojec : P e alence and dis ibu ion o men al diso de s.
Jou nal o Abno mal Psychology, 127(7), 623-638.
{4} Robo ham, D. (2008). S ess among highe educa ion s uden s: Towa ds a esea ch agenda.
Highe Educa ion, 56(6), 735-746.
{5} Bei e , R., Nash, R., McC ady, M., e al. (2015). The p e alence and co ela es o dep ession,
anxie y, and s ess in a sample o college s uden s. Jou nal o A ec i e Diso de s, 173, 90-96.
{6} Has ie, T., Tibshi ani, R., & F iedman, J. (2009). The Elemen s o S a is ical Lea ning: Da a
Mining, In e ence, and P edic ion (2nd ed.). Sp inge Se ies in S a is ics.
{7} Rajkoma , A., Dean, J., & Kohane, I. (2019). Machine lea ning in medicine. New England
Jou nal o Medicine, 380(14), 1347-1358.
{8} Beam, A. L., & Kohane, I. S. (2018). Big da a and machine lea ning in heal h ca e. JAMA,
319(13), 1317-1318.
{9} Ambale-Venka esh, B., Yoneyama, K., Yang, X., e al. (2017). Ca diac MRI albuminu ia in
he elde ly: Algo i hmic quan i ica ion o imp o ed enal unc ion assessmen . Radiology,
281(2), 387-396.
{10} Rajkoma , A., O en, E., Chen, K., e al. (2018). Scalable and accu a e deep lea ning wi h
elec onic heal h eco ds. NPJ Digi al Medicine, 1(1), 18.
{11} Ha che , L., & P us, J. S. (1991). An academic s ess scale de elopmen and psychome ic
analyses. Jou nal o Educa ion and Psychology, 83(1), 113-118.
