T a icIQ: The T a ic Dilemma
Syed Ali Asgha Naq i (24089431)
20 Ma ch 2025
Abs ac
U ban a ic conges ion is a pe ennial issue in ci ies wo ldwide, causing delays, pol-
lu ion, s ess, and sa e y isks o oad use s and pedes ians. This s udy p oposes T a -
icIQ’s AI-based sma signal sys em o add ess hese issues. The sys em p edic s u u e
a ic olume and equency using his o ical da a and dynamically adjus s signal iming
o op imize ehicle low. I also moni o s mino acciden s and pedes ian a ic o educe
acciden s and unnecessa y delays. The esea ch in ol es a comp ehensi e analysis o cu -
en li e a u e, e iewing cu en a ic managemen ini ia i es and in as uc u e, policy,
and echnology up ake gaps. Me hodology in ol es phased ollou , da a-d i en model-
ing, and s akeholde coo dina ion o enable ealis ic in eg a ion in ci y cen e s. Findings
demons a e ha sma signals can g ea ly impac a ic e iciency, enhance sa e y o
pedes ians, and help mo e eme gency ehicles. I s po en ial as a sus ainable, scalable
in elligen solu ion o u ban mobili y and sma conges ion managemen is unde sco ed
by he indings.
1 In oduc ion
In an a icle by INRIX [5], he op 3 ci ies in e ms o a ic conges ion we e Is anbul, New
Yo k and Chicago. Expanding oad lanes alone does no sol e he unde lying p oblem. The e
is a need o a pe manen , solid solu ion. Tha is he goal o ou esea ch and he sole p oblem
he company, T a icIQ is ying o sol e.
Acco ding o a esea ch conduc ed by Anki [3], and his co-au ho s, while discussing he
cause o sudden a ic jams, concluded ha chao ic d i ing bu mo e impo an ly mismanage-
men is he leading cause o in ense a ic jams in ci ies and his necessi a es imp o ed a ic
managemen s a egies. Now ci ies in coun ies like India, Pakis an o Bangladesh whe e he
lack o in as uc u e enabling be e managemen is wo isome ye unde s andable. Howe e ,
i is also a p oblem o big ci ies whe he hey a e in Eu ope o Ame ica.
In a discussion by TimeOu [6], i was highligh ed ha e en ci ies like Be lin, Wa saw and
B ussels ace a ic conges ion and acco ding o he ecen numbe s London was a he op o
he lis o being he mos conges ed ci y in Eu ope. Popula ion densi y, limi ed oad capaci y
and a ic mismanagemen when combined play a signi ican ole in his conges ion. So
ega dless o he coun y o he in as uc u e o public anspo o he ci y, a ic conges ion
is always a p oblem in mos cases.
Hence, he scope o his p ojec unde aken by T a ic IQ is o mi iga e he p oblem o
a ic conges ion. I shall be de e mining ha by p oposing an AI-powe ed solu ion which
we call as sma signals. The sma signals ha e he capabili y o de ec ing he incoming
a ic and p edic ing he olume and equency o ehicles based on his o ical da a esul ing
in an in as uc u e whe e a ic lows eely in addi ion o ha he second cause o a ic
conges ion is also add essed which is mino acciden s which a e o en caused by pedes ians
and such. The e o e, ou sma signals no only p opose o o e a smoo h low o a ic bu
a smoo h low o pedes ians as well.
1
The e o e, o explo e mo e deeply and p opose a easible solu ion o his p oblem his
pape is di ided in o 6 pa s which a e he li e a u e e iew, he de ini ion o he p oblem,
me hodology, echnology selec ion, challenges, implemen a ion plan and inally he conclusion.
2 Li e a u e Re iew
Re iew exis ing li e a u e on AI adop ion in business, challenges, and bes p ac ices. Discuss
ele an case s udies o examples. Iden i y gaps in cu en knowledge ha you p ojec
add esses. You should aim o a leas 10 academic e e ences o be p ope ly ci ed in he inal
pape
In o de o in oduce a sma sys em we i s need o analyze he in as uc u e. Oyewale
[1] Found he basis o in as uc u e o such sma sys ems o be in eg a ed in o a sma ci y.
The gis o he esea ch was he p ecipi a ion o a ic managemen s akeholde s and hei
accep ance o he adop ion o hyb id AI echniques. Th ough ce ain su eys and u he
esea ch i was concluded ha pe cei ed ease o use has a posi i e e ec on a i ude o use.
This ansla es o i some hing is ha d o use hen people will esis o ejec i . In ou case,
we a e adop ing a simila app oach, p e e ences o au onomy ejec s he idea o esis ance. In
ano he s udy, Nasim [8], explo es he need o AI echnology o be in eg a ed in o he a ic
sys em poin ing ou he ac s ha inc ease in popula ion, globaliza ion and mismanagemen
leads o a ic conges ion which in u n leads o pollu ion, s ess and p oduc i i y loss.The
au ho no es ha he me e po en ial o AI o manage a ic wi hou any kind o ex e nal
o ces like human in e en ion is unexplo ed al hough such sys ems and AI models a e being
de eloped. Howe e , he pape d aws a en ion o he challenges aced by AI applica ions
ega dless o he p omise o alle ia e conges ion and imp o e a ic low.
Mo eo e , Pio [11] while s udying he p oblems wi h implemen a ion o sus ainable u -
ban mobili y ad oca ed he ac ha inc ease in he numbe o p i a e ca s combined wi h
a de eloping oad ne wo k is esul ing in an u ban sp awl which in u n esul s in a ic
conges ion. This is one o he easons o i . Al hough done on a na ional le el his s udy
s ill p o ides aluable inpu o he oo o he p oblem o why a ic conges ion is such a big
issue and wha is causing i .
3 P oblem S a emen
As men ioned p e iously, a ic jams a e inc easing in u ban a eas wi hou any solid solu ion
obse ed in he dis an u u e. The a ic jam is he oo o all kinds o p oblems om
pollu ion o oadside inciden s o a hec ic u ban en i onmen c ea ing s ess and panic.
In addi ion o ha , he e a e long and useless wai ing imes o pedes ians which esul in
jumping o signals and ha esul s in un o una e mishaps, simila ly eme gency ehicles ha e
no p io i y when i comes o he ac ual managemen o a ic hey ely on he d i e in on
o make way o hem. Hence, hese a e he main and c i ical p oblems when combined o m
a huge chaos on u ban oads and also a e he p oblems he p oposed solu ion will add ess.
This p ojec aims o a ge a numbe o gaps and oids in cu en knowledge as well as a ge
ma ke niches ha a e no as sa u a ed. As p e iously implemen ed, inc easing he lanes
won’ ha e any e ec on he a ic jam.We p opose an AI-powe ed solu ion so a ic in u ban
a eas can be moni o ed and can help municipali ies o op imize a ic low.
The co e objec i e o he analysis is o help mi iga e o educe a ic in big ci ies. The
p ima y cause o a ic jams is a ic o e load. Cu en ly aiming o come up wi h a solu ion
in Sou h-Eas Asia because o shee popula ion and ex eme a ic jams. Fi s ly, we a e
p oposing a comp ehensi e AI-powe ed solu ion ha me ges a ic low along wi h pedes ian
2
c ossing and sa e y ocusing on mul iple aspec s o u ban mobili y. Secondly, ou aim is o
explo e machine lea ning echniques o de elop a sys em which is e ec i e, sophis ica ed and
obus hence p o iding aluable insigh s in o such sys ems o u u e wo ks.
4 Me hodology
The ollowing sec ion co e s he use case scena io, he echnology selec ion and easibili y
ollowed by he implemen a ion plan.
4.1 Use Case O e iew
AI can be pi o al in a ic managemen by p o iding da a d i en in elligen solu ions.These
a e he key objec i es o he p oposed solu ion
•Dynamic T a ic Signal Con ol: Th ough eal- ime a ic low da a, a ic signal
imings a e adjus ed. Signal changes a in e sec ions a e op imized o ensu e a smoo h
low and a oid majo a ic jams. This is done by using his o ical da a o p edic
pa e ns ha may o m due o ac o s such as peak hou s.
•Inciden Managemen : Roadside inciden s o en lead o a ic jams due o panic o
cu iosi y om he public, o ming bo lenecks. I any oadside inciden s o anomalies
occu , hey a e p omp ly epo ed o he ele an au ho i ies. T a ic signals a e hen
managed dynamically o ensu e ha he a ic low emains una ec ed.
•T a ic Demand Fo ecas ing: By analyzing his o ical da a, local e en s, and wea he
pa e ns, AI can p edic peak a ic pe iods and gene a e essen ial o ecas s. This
p edic i e capabili y is a key objec i e o he p oposed solu ion.
•Real- ime T a ic Flow P edic ion: Real- ime da a om a ic senso s, came as,
and GPS de ices is analyzed o p o ide insigh s o ou e planning. This enables d i e s
o choose op imal ou es, he eby educing conges ion and a el ime. I also helps
p i a e ca owne s a oid high- a ic a eas in ad ance.
4.2 Technology Selec ion and Jus i ica ion
We a e p oposing a machine-lea ning-based solu ion which is ained o manage a ic au-
onomously. Li [7] used an ensemble o machine lea ning me hods o app oxima ely he
same pu pose bu in hei s udy hey used SVM, andom o es and AdaBoos o pa e n
ecogni ion and hei p ima y goal was ehicle classi ica ion.
Hence he use o Machine lea ning me hodology o de ec anomalies and pa e ns is p o en
o be e ec i e. Hence, he eason we also op ed o Machine Lea ning Algo i hm Random
Fo es o ou use case. Random Fo es was speci ically used because i is an ensemble me hod
combining a numbe o decision ees. Addi ionally, i is able o handle high-dimensional da a
which can cap u e complex ela ionships be ween ea u es.
Las ly, a ic pa e ns a e mos ly non-linea hence andom o es ’s abili y o cap u e such
pa e ns is also one o he easons we chose his.Consequen ly, ea u e impo ance measu e
and obus ness make his he pe ec echnique o ou solu ion.
4.3 Scalabili y and Technological Feasibili y
In a pos by Nx Li e [9], hey discuss he cha ac e is ics o andom o es and poin ou he
ac ha Random Fo es can handle housands o inpu a iables wi hou he need o a iable
3
dele ion. This means ha i is pe ec o la ge scale p ojec s whe e huge amoun s o da a is
s o ed and p ocessed.
In addi ion o ha , he Random Fo es algo i hm is ai ly easy o pa allelize as i uns
mul iple decision ees. This is also he eason i is a pe ec i o scalabili y and he e e
inc easing a ic ne wo k. This is an impo an poin when i comes o he cos s uc u e
and he deploymen on la ge scales. As we ha e o make use o he in as uc u e al eady
implemen ed wi h minimum addi ions.
We a e awa e ha mos ci ies o example Ba celona, Islamabad and Mumbai ha e came as
ins alled a e e y a ic signal. Bu h ough esea ch and pe sonal expe ience he ins alla ion
o hese came as ook a signi ican amoun o in es men om local go e nmen s bu p o ed
o be no as e icien as hei main ocus was o egula e a ic laws.
As pe an a icle o CCTV ON RENT [4], Mumbai’s oad su eillance has a huge po en ial
which is unexplo ed. As pe my pe sonal expe ience in Islamabad he capi al o Pakis an and
po ayed as a SAFE CITY [13] he a ic signals we e mainly ins alled o egula e a ic
iola ions bu ail o do e en ha as all a ic iola ions a e penalized by physical a ic
wa dens. Hence, da a om hese sys ems is openly a ailable and we can use he al eady used
in as uc u e o e o i in ou solu ion and no wo y abou ha dwa e compa ibili y.
Random o es s we e conside ed a black box bu due o ecen echnological ad ance-
men s speci ically Shapley Addi i e explana ions (SHAP) which can p o ide insigh s in o he
decision-making p ocess o he model and ad oca ing he ac o he Explainable AI po en ial.
4.4 Challenges in AI Adop ion
The e a e always echnical and non- echnical challenges in such la ge scale p ojec s, some o
hem a e an icipa ed and p ojec ed o coun e while he o he s, you don’ see coming. The e
we e se e al challenges aced by he eam howe e he mos complex ones a e discussed and
a e as ollow,
•Da a Quali y Issues: One o he i s an icipa ed challenges was he inconsis ency
and low quali y o da a om legacy in as uc u e. Much o he collec ed da a was
inconsis en , and lacked cla i y due o which i was no sui able o aining mode n AI
models. To add ess his, a signi ican da a cleaning e o was equi ed, and new de ices
had o be ins alled o ga he highe -quali y and mo e e sa ile da a. This se up ook
almos a mon h, causing delays.
Ve sa ili y was essen ial, as biased da a could lead o biased ou comes. New came as
we e ins alled a selec ed in e sec ions in alpha es ing si es o o e come limi a ions
in olde equipmen . As no ed by Alam [2] om he Uni e si y o he Cumbe lands,
his o ical end analysis o en ails o adap o eal- ime condi ions and canno p edic
anomalies due o i s eliance on ou da ed da a.
A ela ed case s udy om he Los Angeles Sma Ci y ini ia i e showed ha ini ial
ine iciencies we e caused by incompa ible legacy sys ems, and in eg a ing hese sys ems
equi ed addi ional in es men o imp o e accu acy.
•T aining Challenges: Acco ding o Poonam [11], AI a ic managemen sys ems
demand expe ise in compu e ision, machine lea ning, and p edic i e analysis—skills
ha many municipali ies lack. This was also ue in ou case. Al hough an icipa ed, he
apid pace o p ojec de elopmen le us unde p epa ed in e ms o skilled manpowe .
An upskilling p og am became essen ial. As highligh ed by Numalis [?], app oxima ely
43% o o ganiza ions ace AI alen sho ages, o en esul ing in p ojec delays. Many
sma ci y p ojec s ha e only succeeded a e p io i izing pe sonnel aining.
4
•P i acy and Secu i y Conce ns: This eme ged as he mos signi ican challenge.
Public dis us in AI, especially ega ding eal- ime su eillance, aised s ong conce ns.
Moni o ing ac i i ies—whe he o da a collec ion o implemen a ion—spa ked deba es
on indi idual p i acy.
Fo example, ReS ack [12] epo s ha he deploymen o AI-powe ed su eillance cam-
e as in New Yo k Ci y’s subway sys em, in ended o de ec unpaid en ies, igni ed p i-
acy deba es despi e i s in ended secu i y pu pose. These conce ns mus be add essed
wi h anspa ency and s ong da a go e nance policies.
Hence, o coun e his businesses mus ensu e ha hei da a p o ec ion p ac ices o he
da a in s o age and in ansi a e up o he indus y s anda ds and being anspa en o he
gene al public in such ma e s is he bes way o go abou hese si ua ions.
So, in a nu shell ollowing we e he majo challenges ha we e aced by he leade ship
bu imely solu ion o hese challenges made su e ha he obs uc ions don’ gain a lo o
esis ance.
5 Implemen a ion Plan
The implemen a ion plan is di ided in o h ee pa s, he eams, business me ics and ope a-
ional decisions.
5.1 O ganiza ional S uc u e
The o ganiza ional eams a e di ided in o wo pa s: in e nal and ex e nal eams. The
in e nal eam is di ec ly connec ed o he ope a ion o he p ojec and h ee main eams a e
in ol ed, AI and Da a Science Team which consis s o ou AI and ML enginee s esponsible o
de eloping and ine uning he models. Then comes ou Da a Analys s, who a e esponsible
o he da a used o aining and decision making and las ly ou So wa e De elope s which
a e esponsible o in eg a ion o he models wi h he in as uc u e and he de elopmen
ope a ions ha ollow.
The second eam in ol ed in e nally is he IT and In as uc u e Team which consis s o
ou Ne wo k Enginee s esponsible o cloud and edge compu ing amewo k. Ou Cybe se-
cu i y specialis s in cha ge o secu i y and p o ec ion o in ellec ual p ope y and sensi i e
da a om cybe a acks and las ly ou Sys em admins esponsible o adminis a i e asks in
con ex o se e s, da abases and such.
The hi d and inal eam in ol ed in e nally is ou T a ic Ope a ions and Enginee ing
Team. This eam consis s o a ic enginee s p o iding domain expe ise, u ban planne s
ensu ing he implemen a ion o AI is in acco dance wi h he long- e m in as uc u e goals
and ope a ions manage esponsible o supe ising coo dina ion in a ic con ol cen e s.
These a e he eams which a e no di ec ly in ol ed wi h he p ojec bu play a majo ole
in o e all success o he p ojec . I consis s o wo majo eams. The i s one being Policy
and Compliance Team which consis s o ou legal and e hical ad iso s add essing he e hical
and mo al use o AI and has Go e nmen Liaisons esponsible o app o als o egula ions
and such.
Second comes ou Public engagemen and T aining Team. This is also a e y impo an
aspec and consis s o Public Awa eness O ice s esponsible o educa ion o ci izens ega ding
he AI echnology and ga he ing o eedback. Then we ha e Change Managemen Expe s
which ease he public in o AI adop ion and wo k o ce ansi ion and inally ou T aining
Coo dina o s esponsible o up-skilling o s a by de eloping AI aining p og ams.
5
5.2 Business Me ics and KPIs
We a e di iding ou KPIs such ha hey a e measu able and ha e key esponsibili ies o each
ha KPI.
•In e sec ion Delay Reduc ion
Minimize wai imes a a ic signals and imp o e o e all a ic low.
–Goal: Reduce a e age in e sec ion wai imes by 30% wi hin he i s 12 mon hs.
–Achie e a 15% educ ion in wai ime du ing he i s 2 qua e s.
–Ensu e 90% o a ic ligh s dynamically adjus based on conges ion.
–Inc ease h oughpu by 20% wi h no bo lenecks.
•T a ic Volume Managemen
Op imize oad usage by balancing a ic loads ac oss in e sec ions.
–Goal: Main ain a e age oad occupancy below 85% du ing peak hou s.
–Imp o e signal iming accu acy o handle 95% o a ic luc ua ions.
–Reduce lane unde u iliza ion by 25% using signal p io i iza ion.
•Inciden Response E iciency
Ensu e as e clea ance o eme gency ehicles and be e conges ion managemen .
–Goal: Reduce eme gency ehicle esponse ime by 40%.
–Imp o e e ou ing e iciency by 35% o conges ion educ ion.
–Ensu e 80% o signals p io i ize eme gency ehicles wi hin 5 seconds o de ec ion.
•Main enance Cos Reduc ion
Lowe ope a ional expenses associa ed wi h a ic in as uc u e.
–Goal: Dec ease main enance cos s by 25%.
–Reduce human in e en ion by 50%.
–Ex end ha dwa e li espan by 30%.
•Adop ion Ra e & Public Accep ance
Ensu e high public accep ance and usabili y o he sys em.
–Goal: Achie e 75% posi i e eedback wi hin he i s 12 mon hs.
–Launch 3 public awa eness campaigns.
–Reach a 50% adop ion a e ac oss me opoli an a eas.
5.3 Ope a ionaliza ion
The oadmap o in eg a ing AI in o he ope a ion is di ided in o in eg a ion oadmap and
Change managemen S a egies. Fi s we shall discuss he in eg a ion oadmap.
•Feasibili y Analysis and App o als
In he i s s age, he goal is o conduc a easibili y analysis o de e mine whe he
he p ojec ed ou comes a e iable. This in ol es echnical, inancial, and legal assess-
men s. The p ima y objec i e is o secu e app o als om local au ho i ies and ensu e
compliance wi h local a ic egula ions.
6
•De elopmen S age
Once p e equisi es a e ul illed, he p ojec en e s a 9-mon h de elopmen phase. Du ing
his s age, AI models a e ained and ine- uned, and he necessa y in as uc u e o
alpha es ing is se up. All c i ical sys em componen s a e de eloped and p epa ed o
es ing.
•Pilo Deploymen
A e de elopmen , pilo deploymen is ini ia ed using A/B es ing. High- olume in-
e sec ions a e selec ed o deploy he sys em, moni o i s eal- ime pe o mance, and
collec use and sys em eedback o i e a i e imp o emen s.
•AI Model Re inemen
Based on eedback om pilo es ing, hype pa ame e s a e adjus ed and he models a e
u he ine- uned. Con inuous imp o emen is suppo ed h ough eedback loops and
enhanced p edic i e modeling o imp o e accu acy and eliabili y.
•La ge-Scale Deploymen
Following success ul es ing and e inemen , he sys em is deployed on a la ge scale
ac oss ci y-wide ne wo ks. This in ol es expanding om pilo zones while u ilizing
exis ing in as uc u e o minimize he need o new ins alla ions.
•Pos -De elopmen Moni o ing
A e ull deploymen , key pe o mance indica o s (KPIs) a e acked and analyzed.
A eal- ime moni o ing dashboa d is implemen ed o enable con inuous pe o mance
acking and epo ing. T anspa ency wi h he public is emphasized h ough open
access o esul s and sys em ope a ions.
The change managemen s a egies a e di ided as ollows,
•Hyb id Supe ision
A he s a i is a compulso y need o human in e en ion when i comes o supe ision
and moni o ing hence we a e using a hyb id model o moni o ou sys ems o ensu e
hey a e wo king as planned. One hing o no e is ha human dependance is dec eased
he e because du ing sys em ailu es he numbe one p io i y is o main ain he balance
be ween human o e sigh and au oma ion. This is done so ha in u u e no human
in e en ion is needed a any poin .
•T aining o S a
We ha e o upskill he s a in e e y shape and o m. Fo his we shall o e aining
p og ams which will be ocused on AI ope a ions, main enance and oubleshoo ing,
keeping in mind ha s a e-o - he-a concep s a e gi en in o de o he s a o be
skill ul no only o he p ojec bu as a p o essional oo.
5.4 Legal F amewo k
To accommoda e AI ech in o he gene al public changes need o be made hence we ha e o
wo k wi h policymake s o do so. Collabo a ion wi h ci y planne s and ele an au ho i ies is
impo an o ensu e ha ou sys em complies wi h e ol ing a ic laws and s anda ds.
Addi ionally, he whole p ojec li e-cycle mus be aligned wi h he GDPR and EU AI ac .
This means ha we mus ensu e ha all pe sonal da a collec ed and p ocessed mus s ic ly
adhe e o he e hical amewo ks such as da a minimiza ion, pu pose limi a ion and use con-
sen . Addi ionally, we mus implemen da a p o ec ion measu es bo h in s o age and ansi
such as enc yp ion, anonymiza ion and secu e s o age. Mo eo e , comple e anspa ency mus
7
be showcased wi h use s abou how hei da a will be used and p o ide use - iendly access
o da a subjec igh s.
The algo i hms mus be explainable, ai and non-disc imina o y as equi ed by he EU
AI Ac . Las ly, we ha e o es ablish a clea go e nance amewo k o o e see da a e hics bo h
h ough in e nal audi ing and ex e nal eams.
6 Technical De ails
The ollowing sec ion includes he desc ip ion o he da a se and he p ocedu e o model
aining and e alua ion.
6.1 Da a Desc ip ion and Visualiza ion
As men ioned ea lie we a e using andom o es s o ou use case. Be o e doing any hing we
ha e o i s analyze and isualize ou da a. Ou da a has 4 columns ep esen ing Times amp,
Junc ion, Vehicles and ID espec i ely. I cap u es he a ic low a mul iple junc ions o e
ime.
Figu e 1: T a ic Flow T ends du ing he day h ough ce ain junc ions
Figu e 1, ep esen s he a ic low ends du ing he day. We can obse e ha ehicle
coun s a y and inc ease o e ime wi h peaks co esponding o ush hou s. The peaks
indica e imes o high a ic olume which assis in iden i ying he c i ical pe iods o a ic
managemen .
Figu e 2, displays a boxplo ep esen ing he dis ibu ion o a ic olumes ac oss di e en
hou s o he day. The cen al box ep esen he inqua ile ange which is whe e he majo i y o
a ic da a alls o each hou . The line ep esen s he median a ic olume and he whiske s
ex end o show he ange o ypical a ic alues. This isualiza ion helps us iden i y he peak
a ic hou s and anomalies ha may equi e special a en ion du ing he da a p epa a ion
and model aining.
Figu e 3, depic s a line cha isualizing he a e age numbe o ehicles passing h ough
di e en junc ions o e a gi en pe iod o ime. Each line co esponding o a di e en colo
ep esen s a di e en junc ion. F om he hca , i is e iden ha he junc ion 1, expe iences
consis en ly highe a ic low, indica ing po en ial conges ion poin s.
8
Figu e 2: Boxplo o T a ic Dis ibu ion Ac oss Di e en Hou s o he Day
Figu e 3: Line Cha o a e age numbe o ehicles pe junc ion
6.2 Model T aining and E alua ion
As men ioned ea lie , o his pa icula applica ion we a e using andom o es , which makes
an ensemble o decision ees, which will be pe ec o ou use-case scena io as i combines
he ou pu o imp o e accu acy and p e en s o e i ing. The i s s ep in ol ed da a P ep o-
cessing. Ini ially, we isualized he da ase (Figu e 1,2 and 3) o iden i y unde lying pa e ns,
ends and anomalies. Signi ican anomalies we e p uned and missing alues we e eplaced
by he mean o he espec i e ea u e column o main ain da a consis ency and educe skew.
Using a hea map we checked he co ela ion be ween ea u es.
9
