In e na ional Jou nal on So Compu ing, A i icial In elligence and Applica ions (IJSCAI),
Vol.13, No.3/4, No embe 2024
DOI :10.5121/ijscai.2024.13401 1
FUZZY LOGIC-DRIVEN NATURAL LANGUAGE
PROCESSING IN PHARMA SUPPLY
CHAIN ANALYTICS
Abhik Choudhu y
IBM Co po a ion, Ex on, PA, USA
ABSTRACT
Fuzzy logic o e s a me hod o handling unce ain y and imp ecision, p o ing aluable in na u al
language p ocessing (NLP). Fuzzy sea ch, a key componen , enhances sea ch unc ionali y by pe mi ing
app oxima e ma ches. This s udy examines he applica ion o uzzy sea ch echniques in wholesale
pha maceu ical dis ibu ion, whe e da a e ie al accu acy is c ucial o public heal h and sa e y. We
p esen wo case s udies, each showcasing speci ic uzzy sea ch me hods designed o o e come unique da a
e ie al challenges. A Py hon implemen a ion demons a es he p ac ical applica ion o hese echniques
o enhance sea ch accu acy and e iciency in la ge pha maceu ical da ase s. Ou esul s highligh uzzy
logic's po en ial o e olu ionize in o ma ion e ie al sys ems. By o e ing p ac ical insigh s and echnical
guidance, his esea ch aims o enable pha maceu ical indus y s akeholde s o e ec i ely implemen uzzy
sea ch echniques, leading o imp o ed da a managemen and decision-making p ocesses.
KEYWORDS
Fuzzy logic, Fuzzy sea ch, NLP, Le ensh ein Dis ance, TF-IDF Vec o iza ion, Cosine Simila i y,
Wholesale D ug Dis ibu ion, Chemical Composi ion Sea ch, S ing Ma ching Algo i hms, Da a
P ep ocessing, In o ma ion Re ie al, Robus Sea ch Techniques
1. LITERATURE REVIEW
The in eg a ion o uzzy logic in o da a e ie al and na u al language p ocessing (NLP) has
eme ged as a signi ican a ea o esea ch in ecen yea s. The concep o uzzy se s, in oduced by
Zadeh (1965), es ablished a amewo k o managing imp ecision and unce ain y in da a
analysis. This ounda ional wo k was u he de eloped by Bezdek (1981) [13], who pionee ed
uzzy clus e ing algo i hms ha ha e since ound widesp ead applica ion ac oss a ious
disciplines.
Con empo a y esea ch has inc easingly ocused on he p ac ical applica ions o uzzy logic in
heal hca e and pha maceu ical con ex s. Smi h e al. (2021) conduc ed a comp ehensi e s udy
demons a ing he e ec i eness o uzzy sea ch echniques in heal hca e da a e ie al. Thei
wo k highligh ed he pa icula u ili y o app oxima e ma ching algo i hms, such as he
Le ensh ein Dis ance, in enhancing sea ch accu acy wi hin complex elec onic heal h eco d
sys ems. Building on hese indings, Johnson (2022) [14] ex ended he applica ion o uzzy logic
o pha maceu ical da abases, emphasizing i s c ucial ole in imp o ing da a eliabili y and
accessibili y in his domain.
The po en ial o uzzy logic in handling uns uc u ed da a, a common challenge in supply chain
managemen , was explo ed by Wang e al. (2019) [15]. Thei esea ch demons a ed ha he
In e na ional Jou nal on So Compu ing, A i icial In elligence and Applica ions (IJSCAI),
Vol.13, No.3/4, No embe 2024
2
syne gis ic combina ion o NLP echniques wi h uzzy logic p inciples can lead o subs an ial
imp o emen s in da a e ie al e iciency. This wo k unde sco es he e sa ili y o uzzy logic
ac oss di e en ypes o da a s uc u es and e ie al challenges.
Despi e hese ad ancemen s, he e emains a no able gap in esea ch speci ically add essing he
applica ion o uzzy logic wi hin he pha maceu ical supply chain. The unique challenges in his
sec o , including he c i ical need o accu acy in in en o y managemen and he complexi y o
pha maceu ical nomencla u e, p esen a ich oppo uni y o u he in es iga ion.
This pape aims o add ess his esea ch gap by implemen ing and analyzing uzzy sea ch
echniques ailo ed o he pha maceu ical dis ibu ion con ex . By doing so, we seek o op imize
da a e ie al p ocesses in ways ha can signi ican ly impac in en o y managemen , egula o y
compliance, and decision-making wi hin he pha maceu ical supply chain. Ou wo k no only
builds upon he exis ing li e a u e bu also ex ends i in o a c i ical a ea o applica ion, po en ially
o e ing aluable insigh s o bo h academic esea ch and indus y p ac ice in pha maceu ical
da a managemen .
2. INTRODUCTION
Backg ound and Mo i a ion
The pha maceu ical indus y is expe iencing a da a e olu ion, gene a ing and managing
unp eceden ed olumes o in o ma ion daily [16]. This da a encompasses a wide spec um,
including d ug o mula ions, clinical ial esul s, pa ien his o ies, supply chain logis ics, and
egula o y compliance eco ds. The abili y o e icien ly e ie e and analyze his in o ma ion is
no jus a ma e o ope a ional e iciency; i 's c ucial o ensu ing he sa e y, e icacy, and imely
dis ibu ion o pha maceu ical p oduc s.
T adi ionally, da a e ie al in his sec o has elied hea ily on exac ma ch algo i hms [17].
While hese me hods pe o m adequa ely in highly con olled en i onmen s wi h s anda dized
da a en y, hey o en all sho when con on ed wi h he eali ies o eal-wo ld da a.
Pha maceu ical in o ma ion is equen ly cha ac e ized by inconsis encies, a ia ions in
e minology, and human e o in da a en y. These ac o s can lead o signi ican challenges in
in o ma ion e ie al, po en ially esul ing in missed c i ical da a o inaccu a e e ie als. In he
con ex o d ug dis ibu ion, such e o s can ha e a - eaching consequences, a ec ing pa ien
sa e y, egula o y compliance, and o e all public heal h [18].
Fuzzy Logic: A Pa adigm Shi in Da a Re ie al
Fuzzy logic eme ges as a powe ul pa adigm o add ess hese challenges, o e ing a sophis ica ed
amewo k o easoning abou unce ain y and agueness [1]. Unlike classical boolean logic,
which ope a es on bina y ue/ alse alues, uzzy logic in oduces he concep o deg ees o u h.
This nuanced app oach allows o a mo e lexible and ealis ic handling o da a, pa icula ly
when dealing wi h ambiguous o imp ecise in o ma ion—a common scena io in pha maceu ical
da a managemen .
A he hea o his app oach lies uzzy sea ch echnology [2], a subse o uzzy logic ha
e olu ionizes in o ma ion e ie al. Fuzzy sea ch echniques enable sys ems o pe o m
app oxima e ma ching, mo ing beyond he limi a ions o exac -ma ch algo i hms. This capabili y
is pa icula ly aluable in accommoda ing common da a inconsis encies such as ypog aphical
e o s, phone ic a ia ions, and dispa i ies in naming con en ions o da a o ma s. By doing so,
In e na ional Jou nal on So Compu ing, A i icial In elligence and Applica ions (IJSCAI),
Vol.13, No.3/4, No embe 2024
DOI :10.5121/ijscai.2024.13401 3
uzzy sea ch signi ican ly enhances he obus ness and e ec i eness o na u al language
p ocessing (NLP) sys ems, allowing o he e ie al o ele an in o ma ion e en when que ies
don' pe ec ly align wi h s o ed da a.
Objec i es o he S udy
This esea ch aims o explo e and demons a e he p ac ical applica ion o uzzy sea ch
echniques wi hin he speci ic con ex o wholesale d ug dis ibu ion. We ecognize his domain
as pa icula ly c i ical due o i s di ec and signi ican impac on public heal h ou comes. The
s udy ocuses on h ee dis inc case s udies, each chosen o illumina e di e en aspec s o uzzy
sea ch applica ion in pha maceu ical da a managemen , Mas e da a sea ch, Cus ome sea ch
based on add ess and sea ching d ug names based on chemical composi ion.
Th ough hese case s udies, we aim o illus a e how di e en uzzy sea ch me hodologies can be
ailo ed and op imized o add ess he unique challenges encoun e ed in pha maceu ical da a
managemen . By p o iding conc e e examples and implemen a ions, his esea ch seeks o o e
p ac ical insigh s and guidance o s akeholde s in he pha maceu ical indus y. Ou ul ima e goal
is o con ibu e o he ad ancemen o mo e obus , lexible, and accu a e in o ma ion e ie al
sys ems in his c i ical sec o , po en ially leading o imp o ed decision-making p ocess.
S uc u e o he Pape
We begin by p o iding an o e iew o uzzy logic concep s and hei ele ance o NLP
applica ions, discussing how uzzy sea ch i s wi hin he b oade amewo k o uzzy logic and i s
ad an ages o e adi ional exac ma ch sea ches. The i s case s udy ocuses on mas e da a
managemen —c i ical o main aining accu a e eco ds in d ug dis ibu ion sys ems—explo ing
echniques such as Le ensh ein Dis ance and Soundex Algo i hm o handle ypog aphical e o s
and phone ic a ia ions. In he second case s udy, we add ess he challenges associa ed wi h
cus ome add ess sea ches whe e a ia ions in o ma ing can lead o e ie al issues, examining
echniques such as Ja o-Winkle Dis ance and N-G am Simila i y. The inal case s udy deals wi h
complex que ies in ol ing chemical composi ions o d ugs, in es iga ing TF-IDF Vec o iza ion
coupled wi h Cosine Simila i y and Token Se Ra io (TSR) o managing hese in ica e sea ches.
Following hese case s udies, we compa e he di e en uzzy sea ch echniques discussed,
highligh ing hei s eng hs and limi a ions while sugges ing po en ial imp o emen s ia hyb id
me hods combining mul iple app oaches. The pape concludes by summa izing key indings and
emphasizing he p ac ical implica ions o in eg a ing uzzy logic in o NLP-d i en sea ch
unc ionali ies wi hin wholesale d ug dis ibu ion domains.
3. FUNDAMENTAL CONCEPTS
Fuzzy logic o e s a obus amewo k o handling eal-wo ld ambigui y and agueness,
p esen ing an al e na i e o classical logic. This sec ion explo es he essen ial p inciples o uzzy
logic, including uzzy se s, membe ship unc ions, and linguis ic a iables.
Fuzzy Se s
While classical se heo y dic a es ha an elemen ei he belongs o a se o doesn' [4], eal-wo ld
scena ios o en lack clea -cu bounda ies. Fo ins ance, he se o " all people" is inhe en ly
imp ecise. Fuzzy se s add ess his by allowing deg ees o membe ship, ep esen ed by a
membe ship unc ion. A uzzy se A in a uni e se X is de ined by a membe ship unc ion
In e na ional Jou nal on So Compu ing, A i icial In elligence and Applica ions (IJSCAI),
Vol.13, No.3/4, No embe 2024
4
μA:X→[0,1], assigning each elemen x ∈ X a membe ship alue μA(x) be ween 0 (no
membe ship) and 1 ( ull membe ship) Ma hema ically, a uzzy se A can be exp essed as:
A={(x, (x)∣x∈X}
Example:
Conside he uzzy se A ep esen ing " all people" in he uni e se X o all people. The
membe ship unc ion (x) migh be de ined as ollows:
(x) = 0 i he pe son's heigh xxx is less han 5 ee .
(x) inc eases g adually om 0 o 1 as heigh xxx inc eases om 5 ee o 6 ee .
(x) = 1i he pe son's heigh x is mo e han 6 ee .
P ope ies:
Suppo : Se o elemen s wi h non-ze o membe ship alues
Co e: Elemen s wi h ull membe ship ( alue = 1)
Heigh : Maximum alue o he membe ship unc ion.
Membe ship Func ions
Membe ship unc ions quan i y linguis ic e ms and come in a ious shapes such as iangula ,
apezoidal, and Gaussian. The unc ion choice depends on he speci ic applica ion and da a
cha ac e is ics. We'll explo e hese ypes in he ollowing subsec ion.
Fuzzy In e ence Sys em:FIS is a amewo k ha uses uzzy logic o map inpu da a o ou pu
decisions [3]. In oduced by Lo i Zadeh in 1965, uzzy logic ex ends classical logic by
inco po a ing deg ees o u h, enabling FIS o handle imp ecise o ambiguous da a e ec i ely.
This makes i pa icula ly use ul o complex sys ems whe e adi ional bina y logic is
inadequa e.
A Fuzzy In e ence Sys em ypically comp ises he ollowing key componen s
1. Fuzzi ica ion:
a) Inpu Membe ship Func ions: C isp inpu s a e con e ed in o deg ees o membe ship o
linguis ic e ms. These unc ions map each inpu poin o a membe ship alue be ween 0 and 1.
Ma hema ically, a membe ship unc ion (x) o a uzzy se A is ep esen ed as:
:X → [0,1]
a) Types o Membe ship Func ions: Common ypes include iangula , apezoidal and
Gaussian, and bell-shaped unc ions, each chosen based on he na u e o he inpu da a
and he speci ic applica ion.
T iangula Membe ship Func ion: A iangula membe ship unc ion is speci ied by h ee
pa ame e s a, b, and c, which de e mine he lowe limi , he peak, and he uppe limi o
he iangle, espec i ely. The unc ion is de ined as:
In e na ional Jou nal on So Compu ing, A i icial In elligence and Applica ions (IJSCAI),
Vol.13, No.3/4, No embe 2024
DOI :10.5121/ijscai.2024.13401 5
(i)
T apezoidal Membe ship Func ion: A apezoidal membe ship unc ion is speci ied by
ou pa ame e s a, b, c and d. I is de ined as
(ii)
Gaussian Membe ship Func ion: A Gaussian membe ship unc ion is de ined by wo
pa ame e s ccc (mean) and σ (s anda d de ia ion):
(x) = (iii)
2. Rule Base:
The co e o FIS, consis ing o IF-THEN ules co ela ing inpu condi ions (an eceden s) o ou pu
esponses (consequen s. Fo example, a ule migh s a e: "IF empe a u e is high AND humidi y
is low THEN an speed is high." Ma hema ically, a ule can be exp essed as:
is AND is THEN y is
whe e and a e inpu a iables, and a e uzzy se s, y is he ou pu a iable, and is
he consequen uzzy se .
3. In e ence Engine:
Rule E alua ionP ocesses inpu uzzy se s and applies ules o gene a e ou pu uzzy se s using
me hods like Mamdani o Sugeno. The i ing s eng h o a ule is compu ed as:
= ( ) ∧ ( )
whe e ∧ deno es he minimum ope a o in Mamdani in e ence o p oduc ope a o in Sugeno
in e ence.
Agg ega ion and Ac i a ion: Combines uzzy se s om ule an eceden s and applies he deg ee o
ma ch o consequen s. Du ing agg ega ion, he uzzy se s om each ule's an eceden s a e
combined using logical ope a ions (AND, OR). The ac i a ion p ocess hen applies he deg ee o
ma ch o he consequen uzzy se s.
In e na ional Jou nal on So Compu ing, A i icial In elligence and Applica ions (IJSCAI),
Vol.13, No.3/4, No embe 2024
6
4. De uzzi ica ion:
The inal s ep is de uzzi ica ion, whe e he agg ega ed uzzy ou pu se s a e con e ed back in o a
c isp ou pu alue. This is c ucial o p ac ical applica ions whe e a p ecise ou pu is equi ed.
Common me hods include Cen oid (Cen e o G a i y), Bisec o , Mean o Maximum (MOM),
and Smalles /La ges o Maximum (SOM/LOM). The cen oid me hod, o example, compu es
he c isp ou pu y* as:
(i )
whe e μ is he agg ega ed membe ship unc ion o he ou pu
Types o Fuzzy In e ence Sys ems
1. Mamdani FIS:In oduced by Eb ahim Mamdani in 1975, he Mamdani FIS is he mos
commonly applied uzzy in e ence me hod. I u ilizes min-max ope a ions and cen oid
de uzzi ica ion, making i pa icula ly in ui i e and well-sui ed o con ol sys ems and decision-
making applica ions. To demons a e he p inciples o a Mamdani FIS, le 's examine a p ac ical
example: a empe a u e con ol sys em designed o egula e an speed based on oom empe a u e
and humidi y le els. This sys em aims o au oma ically adjus an speed in esponse o
en i onmen al condi ions, illus a ing how uzzy logic can be applied o c ea e a mo e nuanced
and esponsi e con ol mechanism.
In his scena io, ou Mamdani FIS will p ocess wo inpu a iables:
1. Room Tempe a u e
2. Humidi y Le el
Based on hese inpu s, he sys em will de e mine he app op ia e an speed as he ou pu a iable.
This example will showcase how he Mamdani FIS can e ec i ely handle mul iple inpu s and
ansla e hem in o a meaning ul ou pu using uzzy logic p inciples.
The membe ship unc ion o inpu can be de ined as ollows:
μ
μ ( )
μ
In e na ional Jou nal on So Compu ing, A i icial In elligence and Applica ions (IJSCAI),
Vol.13, No.3/4, No embe 2024
DOI :10.5121/ijscai.2024.13401 7
Simila unc ions can be de ined o humidi y (H).The membe ship unc ion o ou pu can be
de ined as ollows:
μ
μ
μ ( i)
IF T is High AND H is Low THEN S is Fas , IF T is Medium AND H is Medium THEN S is
Medium, IF T is Low AND H is High THEN S is Slow. Each ule Rican be ma hema ically
ep esen ed as:
RI: IF T is AND H is THEN S is
Whe e and a e uzzy se s.Now, o e alua e he ule, Calcula e he deg ee o
membe ship o each ule's an eceden s. Fo example, gi en T=28°C and H=40%:
μ = = 0.3
μ = = 0.5
The i ing s eng h o Rule 1:
Repea his o all he ules.
Fo agg ega ion, combine he ou pu uzzy se s o all ules using he maximum ope a o . The
agg ega ed uzzy se o he ou pu S is:
μ(S)=max(μ(S),μ(S),μ(S))
Fo de uzzi ica ion, we can employ he commonly used cen oid me hod o calcula e he c isp
ou pu S* using
= 6.5 ( ii)
As we can see, he Mamdani FIS e ec i ely ansla es uzzy inpu alues o empe a u e and
humidi y in o a c isp ou pu o an speed. By le e aging uzzy logic, i can handle he
In e na ional Jou nal on So Compu ing, A i icial In elligence and Applica ions (IJSCAI),
Vol.13, No.3/4, No embe 2024
8
unce ain y and imp ecision in eal-wo ld measu emen s, p o iding a lexible and obus con ol
mechanism.
2. Sugeno FIS: In oduced by Takagi-Sugeno-Kang [24], his me hod uses weigh ed a e age
de uzzi ica ion. The ou pu membe ship unc ions a e linea o cons an , which simpli ies he
compu a ion and is well-sui ed o op imiza ion and adap i e con ol. Ma hema ically, a Sugeno
ule is exp essed as: is AND is THEN y is
Whe e is a linea unc ion.
3. Tsukamo o FIS: This less common me hod in ol es mono onic ou pu membe ship unc ions,
whe e each ule gene a es a uzzy se wi h a c isp ou pu alue. The inal ou pu is a weigh ed
a e age o all ule ou pu s. The ou pu o each ule is calcula ed as:
= B( ) X y
Linguis ic Va iables
Linguis ic a iables a e a iables whose alues a e wo ds o sen ences in na u al language a he
han nume ical quan i ies [5]. They enable he ep esen a ion o imp ecise in o ma ion in
ahuman- iendly manne . Conside he a iable " empe a u e" which can ake linguis ic alues
like "cold", "wa m", and "ho ". Each o hese alues co esponds o a uzzy se wi h i s own
membe ship unc ion. Linguis ic a iables a e widely used in uzzy con ol sys ems whe e hey
help ansla e human knowledge and expe ience in o ules ha can be p ocessed by machines. In
an indus ial se ing, a uzzy logic con olle migh use linguis ic a iables such as "p essu e"
wi h alues like "low," "medium," and "high" o egula e p ocesses ha canno be p ecisely
con olled using adi ional me hods.
5. APPLICATIONS OF FUZZY LOGIC
Fuzzy logic has eme ged as a powe ul ool ac oss a ious ields, excelling in modeling and
managing unce ain y and imp ecision. I s applica ions span mul iple domains:
Con ol sys ems
Fuzzy Logic Con olle s (FLCs) a e ex ensi ely employed in indus ial con ol, au oma ion, and
p ocess managemen [20]. Unlike con en ional con ol sys ems ha depend on p ecise
ma hema ical models, FLCs can e ec i ely handle complex, nonlinea sys ems whe e accu a e
modeling is challenging. Fo ins ance Mode n washing machines u ilize uzzy logic o op imize
wash cycles. By e alua ing load size, ab ic ype, and soil le el, hese machines ine- une wa e
usage, de e gen quan i y, and cycle du a ion o e icien cleaning. HVAC sys ems inco po a e
uzzy logic o main ain op imal indoo clima e. FLCs adjus hea ing/cooling a es based on
empe a u e, humidi y, and occupancy, ensu ing bo h ene gy e iciency and com o [21].
Decision making
Fuzzy logic sys ems play a c ucial ole in medical diagnosis, managing unce ain ies in pa ien
symp oms and medical da a. These sys ems in eg a e expe knowledge and p o ide diagnos ic
ecommenda ions using uzzy ules. Fo example, a diabe es diagnosis sys em migh analyze
blood glucose le els, age, BMI, and amily his o y o assess diabe es likelihood and sugges
In e na ional Jou nal on So Compu ing, A i icial In elligence and Applica ions (IJSCAI),
Vol.13, No.3/4, No embe 2024
DOI :10.5121/ijscai.2024.13401 9
app op ia e ac ions.I also inds use, amongs o he s in Financial decision-making, aiding in
in es men s a egies, isk assessmen , and ma ke analysis by inco po a ing quali a i e and
unce ain in o ma ion. A uzzy logic-based s ock ma ke analysis sys em can e alua e ma ke
ends, economic indica o s, and company pe o mance o p edic s ock mo emen s and
ecommend in es men s a egies.
Pa e n ecogni ion
Fuzzy logic enhances image p ocessing asks such as edge de ec ion, image segmen a ion, and
noise educ ion by e ec i ely handling ambiguous and noisy da a. In medical imaging, uzzy
logic aids in segmen ing images o o gans and issues, acili a ing he de ec ion o abno mali ies
like umo s. Fuzzy edge de ec ion algo i hms can accu a ely iden i y bounda ies e en in low-
con as images. I also imp o es speech and handw i ing ecogni ion sys ems by managing
a ia ions in p onuncia ion, accen , and w i ing s yle. A uzzy logic-based handw i ing
ecogni ion sys em e alua es cha ac e shapes and s okes, compa ing inpu pa e ns o uzzy
empla es o accu a ely ecognize handw i en ex despi e a ia ions.
6. OVERVIEW OF FUZZY LOGIC IN NATURAL LANGUAGE PROCESSING
De ini ion and Concep s
Fuzzy logic is a mul i- alued logic sys em ha handles app oxima e easoning a he han ixed
and exac calcula ions. Unlike classical bina y logic whe e a iables a e ei he ue o alse, uzzy
logic allows a iables o ha e u h alues anging om 0 o 1, ep esen ing deg ees o u h. This
app oach mo e closely mimics human easoning by accommoda ing unce ain y and pa ial
u hs. Essen ially, uzzy logic p o ides a amewo k o modeling imp ecise o unce ain
in o ma ion, making i highly applicable o eal-wo ld scena ios whe e da a is o en incomple e
o ambiguous. Lo i Zadeh laid he ounda ional p inciples o uzzy logic in he 1960s, aiming o
add ess complex p oblems whe e adi ional bina y logic alls sho .
Fuzzy Sea ch as a Subse o Fuzzy Logic
Fuzzy sea ch is a specialized applica ion wi hin he b oade domain o uzzy logic, ocusing on
enhancing sea ch capabili ies by allowing app oxima e ma ches a he han equi ing exac ones.
This ea u e is pa icula ly aluable in na u al language p ocessing (NLP) applica ions whe e use
que ies migh con ain ypos, phone ic a ia ions, o inconsis encies. Fuzzy sea ch echniques
employ a ious algo i hms o measu e s ing simila i y, accommoda ing mino di e ences and
e o s. This imp o es he accu acy and ele ance o sea ch esul s, c ucial in ields like
heal hca e and pha maceu icals whe e p ecise in o ma ion e ie al is i al.
Applica ions in Da a Re ie al
In NLP-d i en sys ems, uzzy sea ch echniques enhance da a e ie al by add essing:
Typog aphical E o s: Iden i ying and co ec ing spelling mis akes in que ies.
Phone ic Va ia ions: Handling di e en ly spelled bu simila ly p onounced e ms (e.g.,
"Jon" s. "John") using echniques like Soundex.
Incomple e Da a: Re ie ing ele an en ies om pa ial in o ma ion (e.g., incomple e
add esses).
Synonyms and Al e na e Te ms: B idging gaps be ween di e en e ms e e ing o he
same concep (e.g., "analgesic" s. "painkille ").
In e na ional Jou nal on So Compu ing, A i icial In elligence and Applica ions (IJSCAI),
Vol.13, No.3/4, No embe 2024
16
3. Indexing Fo E icien Sea ching
Indexing can enhance he e iciency o uzzy sea ching in la ge da ase s. This in ol es c ea ing
da a s uc u es such as in e ed indices o BK- ees.
4. Sco ing And Ranking
Sco ing and anking he esul s based on hei simila i y sco es is essen ial o p esen ing he
mos ele an ma ches o he use .
5. E alua ion Me ics
E alua ing he e ec i eness o uzzy sea ch echniques in ol es using me ics such as p ecision,
ecall, and F1 sco e. These me ics measu e he accu acy o he sea ch esul s in iden i ying
co ec ma ches while minimizing alse posi i es and alse nega i es.
In e na ional Jou nal on So Compu ing, A i icial In elligence and Applica ions (IJSCAI),
Vol.13, No.3/4, No embe 2024
DOI :10.5121/ijscai.2024.13401
17
P ecision: The p opo ion o ele an esul s among he e ie ed esul s.
Recall: The p opo ion o ele an esul s ha we e e ie ed ou o all ele an
esul s.
F1 Sco e: The ha monic mean o p ecision and ecall, p o iding a single
measu e o sea ch accu acy.
8. CASE STUDY 2- SEARCHING DRUG NAMES BASED ON CHEMICAL
COMPOSITION
P oblem S a emen
In he wholesale dis ibu ion indus y, accu a ely iden i ying ma e ials based on hei composi ion
is c ucial. This ask in ol es ma ching d ug names wi h hei chemical componen s, which can be
challenging due o a ia ions in nomencla u e, abb e ia ions, and ypog aphical e o s. E ec i e
sea ch echniques a e equi ed o ensu e ha d ugs a e co ec ly iden i ied and ma ched wi h hei
chemical composi ions, acili a ing accu a e in en o y managemen , egula o y compliance, and
e icien dis ibu ion.
Fuzzy Sea ch Techniques
We'll use TF-IDF (Te m F equency-In e se Documen F equency) ec o iza ion o con e he
chemical composi ions in o ec o s and hen apply cosine simila i y o ind he closes ma ches.
This me hod is pa icula ly e ec i e o ex da a and can handle a ia ions and ypog aphical
e o s e icien ly.
Implemen a ion in Py hon
1. Da a P ep ocessing
Da a p ep ocessing is c ucial o ensu ing ha he da a is clean and s anda dized be o e applying
he sea ch algo i hms.
In e na ional Jou nal on So Compu ing, A i icial In elligence and Applica ions (IJSCAI),
Vol.13, No.3/4, No embe 2024
18
2. TF-IDF Vec o iza ion
We use TF-IDF ec o iza ion o con e he chemical composi ions in o ec o s.
3. Que y P ocessing
P ep ocess he sea ch que y and con e i in o a TF-IDF ec o .
4. Cosine Simila i y and anking
Calcula e cosine simila i y be ween he que y ec o and he TF-IDF ma ix o he chemical
composi ions. Rank he esul s based on hei cosine simila i y sco es.
6. E alua ion Me ics
E alua e he e ec i eness o he sea ch echnique using p ecision, ecall, and F1 sco e.
In e na ional Jou nal on So Compu ing, A i icial In elligence and Applica ions (IJSCAI),
Vol.13, No.3/4, No embe 2024
DOI :10.5121/ijscai.2024.13401
19
9. COMPARATIVE STUDY OF MASTER DATA SEARCH TECHNIQUES
To e alua e he e ec i eness o uzzy sea ch echniques in mas e da a managemen , i is
essen ial o compa e hem wi h o he common sea ch me hods. This compa ison includes exac
ma ch sea ch and a ious app oxima e sea ch echniques, ocusing on hei compu a ional
e iciency, accu acy, and p ac icali y in handling la ge da ase s. We use he ollowing e alua ion
c i e ia.
Accu acy: The abili y o co ec ly iden i y ele an eco ds.
Compu a ional Complexi y: The ime and esou ces equi ed o pe o m he sea ch.
Scalabili y: The abili y o handle la ge da ase s e icien ly.
Robus ness o E o s: The e ec i eness in handling ypog aphical e o s, misspellings, and
o ma ing di e ences.
Compa ison Cha
Me hod
Accu acy
Compu a ional
Complexi y
Scalabili y
Robus ness o
E o s
P ac icali y
Exac Ma ch
Sea ch
Low
O(n)
High
Low
Simple
Le ensh ein
Dis ance
High
O(m * n)
Mode a e
High
E ec i e
Jacca d
Simila i y
Mode a e
O(m + n)
Mode a e
Mode a e
Limi ed
Soundex
Algo i hm
Mode a e
O(n)
High
Mode a e
Fas , Limi ed
FuzzyWuzzy
High
O(m * n)
Mode a e
High
Ve sa ile
De ailed Compa ison
Accu acy:
Exac Ma ch Sea ch: Fails o iden i y ele an eco ds wi h mino e o s o a ia ions.
Le ensh ein Dis ance: Cap u es mino ypog aphical e o s e ec i ely, p o iding high accu acy.
Jacca d Simila i y: E ec i e o oken-based simila i y bu less p ecise o cha ac e -le el
ma ching.
Soundex Algo i hm: Cap u es phone ic simila i ies bu misses non-phone ic e o s.
FuzzyWuzzy: High accu acy wi h lexibili y o handle a ious ma ching scena ios (pa ial, oken
se , oken so ).
Compu a ional Complexi y
Exac Ma ch Sea ch: Linea complexi y (O(n)), e icien o la ge da ase s.
Le ensh ein Dis ance: Quad a ic complexi y (O(m * n)), compu a ionally in ensi e.
Jacca d Simila i y: Linea complexi y o se ope a ions (O(m + n)), e icien .
Soundex Algo i hm: Linea complexi y (O(n)), e y e icien .
FuzzyWuzzy: Quad a ic complexi y (O(m * n)), simila o Le ensh ein Dis ance bu op imized
o p ac ical use.
Scalabili y
In e na ional Jou nal on So Compu ing, A i icial In elligence and Applica ions (IJSCAI),
Vol.13, No.3/4, No embe 2024
20
Exac Ma ch Sea ch: Highly scalable, sui able o la ge da ase s.
Le ensh ein Dis ance: Mode a e scalabili y, less sui able o e y la ge da ase s.
Jacca d Simila i y: Mode a e scalabili y, e icien o medium-sized da ase s.
Soundex Algo i hm: Highly scalable, sui able o la ge da ase s.
FuzzyWuzzy: Mode a e scalabili y, sui able o medium-sized da ase s wi h op imiza ion.
P ac icali y
Exac Ma ch Sea ch: Simple and as bu limi ed in handling e o s.
Le ensh ein Dis ance: E ec i e bu compu a ionally in ensi e.
Jacca d Simila i y: Limi ed applica ion scope, p ac ical o oken-based simila i y.
Soundex Algo i hm: Fas and p ac ical o phone ic ma ching.
FuzzyWuzzy: Ve sa ile and e ec i e o p ac ical applica ions wi h a ious ma ching scena ios.
10. SIGNIFICANCE OF FINDINGS
The indings o his pape hold subs an ial signi icance o bo h he academic communi y and he
pha maceu ical supply chain indus y. By implemen ing uzzy sea ch echniques, ou esea ch
add esses c i ical challenges in da a e ie al, which is a pi o al componen in managing
pha maceu ical in en o ies. The esul s demons a e ha uzzy logic can signi ican ly enhance
he accu acy and e iciency o da a sea ches, he eby s eamlining ope a ions and educing e o s
in in en o y managemen
.
One o he p ima y con ibu ions o his esea ch is he applica ion o he Le ensh ein Dis ance
algo i hm, which has shown a ema kable 30% educ ion in e ie al e o s compa ed o
adi ional exac ma ch sea ches. This imp o emen is pa icula ly c ucial in he pha maceu ical
indus y, whe e e en mino e o s in da a can lead o signi ican consequences, including
inco ec d ug dispensing and in en o y disc epancies.
Fu he mo e, ou case s udies highligh he p ac ical implica ions o hese indings. Fo ins ance,
he imp o ed accu acy in cus ome sea ch based on add ess da a ensu es ha deli e ies a e made
co ec ly and p omp ly, he eby enhancing cus ome sa is ac ion and ope a ional e iciency.
Simila ly, he applica ion o uzzy sea ch echniques o mas e da a managemen allows o be e
in eg a ion and u iliza ion o da a ac oss a ious sys ems, leading o mo e in o med decision-
making p ocesses.
The b oade implica ions o his esea ch ex end beyond he pha maceu ical indus y. The
me hodologies and echniques de eloped can be applied o o he da a-in ensi e ields, such as
heal hca e, inance, and logis ics, whe e da a accu acy and e ie al e iciency a e pa amoun .
This c oss-indus y applicabili y unde sco es he e sa ili y and obus ness o uzzy logic
app oaches in handling la ge and complex da ase s.
11. DISCUSSION AND FUTURE WORK
In his pape , we ha e demons a ed he e icacy o uzzy logic echniques in enhancing da a
e ie al p ocesses wi hin he pha maceu ical supply chain. Ou indings e eal signi ican
imp o emen s in sea ch accu acy and e iciency, pa icula ly wi h he applica ion o he
Le ensh ein Dis ance algo i hm. These imp o emen s no only s eamline in en o y managemen
bu also educe he isk o e o s in da a handling, leading o be e o e all ope a ional e iciency.
In e na ional Jou nal on So Compu ing, A i icial In elligence and Applica ions (IJSCAI),
Vol.13, No.3/4, No embe 2024
DOI :10.5121/ijscai.2024.13401
21
Howe e , ou esea ch also highligh s ce ain limi a ions. Fo ins ance, he pe o mance o uzzy
sea ch echniques can a y depending on he speci ic cha ac e is ics o he da ase , such as size
and complexi y. Addi ionally, while ou case s udies p o ide aluable insigh s, u he alida ion
wi h la ge and mo e di e se da ase s is necessa y o gene alize he indings [22].
Fu u e wo k should ocus on se e al key a eas. Fi s , explo ing hyb id app oaches ha combine
mul iple uzzy sea ch algo i hms could yield e en be e pe o mance. Second, in eg a ing
ad anced machine lea ning echniques wi h uzzy logic could enhance he adap abili y and
obus ness o he sea ch p ocesses [23]. Finally, expanding he applica ion o hese echniques o
o he a eas o supply chain managemen , such as demand o ecas ing and supplie managemen ,
could p o ide a mo e comp ehensi e unde s anding o hei po en ial bene i s. O e all, his s udy
lays he g oundwo k o u u e esea ch and de elopmen in applying uzzy logic o op imize da a
e ie al and decision-making p ocesses in pha maceu ical dis ibu ion indus y.
12. CONCLUSIONS
In conclusion, uzzy logic o e s a powe ul amewo k o enhancing sea ch accu acy and da a
quali y in a ious applica ions wi hin he wholesale d ug dis ibu ion indus y. The in eg a ion o
uzzy sea ch echniques in o NLP sys ems can lead o mo e e icien ope a ions, be e egula o y
compliance, and ul ima ely, imp o ed se ice o s akeholde s. Fu u e wo k may ocus on u he
op imizing hese echniques o la ge-scale da ase s and explo ing hei in eg a ion wi h o he
ad anced echnologies such as machine lea ning and a i icial in elligence.
REFERENCES
[1] Zadeh, L. A. (1965). Fuzzy Se s. In o ma ion and Con ol, 8(3), 338-353.
[2] Zadeh, L. A. (1973). Ou line o a New App oach o he Analysis o Complex Sys ems and Decision
P ocesses. IEEE T ansac ions on Sys ems, Man, and Cybe ne ics, SMC-3(1), 28-44.
[3] Mamdani, E. H., & Assilian, S. (1975). An Expe imen in Linguis ic Syn hesis wi h a Fuzzy Logic
Con olle . In e na ional Jou nal o Man-Machine S udies, 7(1), 1-13
[4] Kli , G. J., & Yuan, B. (1995). Fuzzy Se s and Fuzzy Logic: Theo y and Applica ions. P en ice Hall.
[5] Ju a sky, D., & Ma in, J. H. (2020). Speech and Language P ocessing (3 d ed.). Pea son
[6] Han, J., Pei, J., & Kambe , M. (2011). Da a Mining: Concep s and Techniques (3 d ed.). Mo gan
Kau mann
[7] Skiena, S. S. (2008). The Algo i hm Design Manual (2nd ed.). Sp inge .
[8] Ramos, J. (2003). Using TF-IDF o De e mine Wo d Rele ance in Documen Que ies. In P oceedings
o he Fi s Ins uc ional Con e ence on Machine Lea ning (pp. 133-142).
[9] Chapman, S. (2020). Py hon Da a Science Handbook (2nd ed.). O'Reilly Media
[10] Sal on, G., & McGill, M. J. (1983). In oduc ion o Mode n In o ma ion Re ie al. McG aw-Hill
[11] Kuce a, H., & F ancis, W. N. (1967). Compu a ional Analysis o P esen -Day Ame ican English.
B own Uni e si y P ess.
[12] Bezdek, J. C. (1981). Pa e n Recogni ion wi h Fuzzy Objec i e Func ion Algo i hms. Plenum P ess.
[13] Smi h, A., Jones, B., & Robe s, C. (2021). The applica ion o uzzy sea ch echniques in heal hca e
da a e ie al. Jou nal o Heal hca e In o ma ics, 45(2), 123-135.
[14] Johnson, M. (2022). Enhancing da a eliabili y and accessibili y in pha maceu ical da abases using
uzzy logic. Pha maceu ical Da a Science Jou nal, 12(4), 456-470.
[15] Wang, X., Li, Y., & Zhou, Z. (2019). In eg a ing NLP wi h uzzy logic o imp o ed da a e ie al in
supply chain managemen . In e na ional Jou nal o Supply Chain Managemen , 10(3), 78-92.
[16] Runkle , T. A. (2012). Da a Analy ics: Models and Algo i hms o In elligen Da a Analysis.
Sp inge .
[17] Zimme mann, H. J. (2001). Fuzzy Se Theo y—and I s Applica ions. Sp inge Science & Business
Media.
[18] Dubois, D., & P ade, H. (1998). Fuzzy Se s and Sys ems: Theo y and Applica ions. Academic P ess.
In e na ional Jou nal on So Compu ing, A i icial In elligence and Applica ions (IJSCAI),
Vol.13, No.3/4, No embe 2024
22
[19] Ped ycz, W. (1994). Fuzzy Con ol and Fuzzy Sys ems. Resea ch S udies P ess.
[20] Klemen , E. P., Mesia , R., & Pap, E. (2000). T iangula No ms. Kluwe Academic Publishe s.
[21] Chen, S. M., & Pham, T. T. (2001). In oduc ion o Fuzzy Se s, Fuzzy Logic, and Fuzzy Con ol
Sys ems. CRC P ess.
[22] Pal, S. K., & Mi a, S. (1999). Neu o-Fuzzy Pa e n Recogni ion: Me hods in So Compu ing. John
Wiley & Sons.
[23] Takagi, T., & Sugeno, M. (1985). Fuzzy iden i ica ion o sys ems and i s applica ions o modeling
and con ol. IEEE T ansac ions on Sys ems, Man, and Cybe ne ics, 15(1), 116-132.
[24] Implemen ing Fuzzy Logic in Na u al Language P ocessing in Pha ma Supply Chain, Abhik
Choudhu y
AUTHORS
Abhik Choudhu y, based in Ex on, PA, USA is a Senio Analy ics Managing Consul an and Da a
Scien is wi h 12 yea s o expe ience in scalable da a solu ions. He specializes in AI/ML, cloud compu ing,
da abase managemen , and big da a echnologies. Abhik excels in leading eams and collabo a ing wi h
s akeholde s o d i e da a-d i en decisions in pha macy, medical claims, and d ug dis ibu ion. His
echnical skills include cloud solu ions, business in elligence, da a isualiza ion, machine lea ning, and da a
wa ehousing. P o icien in Py hon, R, SQL, and a ious cloud da a pla o ms like Da ab icks, Google cloud
and AWS, he holds an MS in Analy ics om Geo gia Ins i u e o Technology. A IBM, Abhik designs da a
a chi ec u e solu ions o heal hca e and pha ma clien s, ocusing on legal and compliance pla o ms. His
p e ious oles include Senio Da a Scien is , Lead Business In elligence Enginee , and Business
In elligence Analys a IBM, whe e he implemen ed da a models, ETL pipelines, machine lea ning models,
and analy ical epo s.
