A Fuzzy Decision Support System with Semantic Knowledge Graph for

Personalized Asthma Monitoring: A Conceptual Modeling

Ayan Chatterjee

Dept. of Digital Technology, STIFTELSEN NILU, 2007 Kjeller, Norway

Keywords:

Fuzzy Ontology, Decision Support System, Asthma Monitoring, Knowledge Representation.

Abstract:

Asthma, a complex chronic respiratory condition, poses signiﬁcant management challenges, necessitating

personalized monitoring for optimal treatment outcomes and individual well-being. This study introduces a

Fuzzy Decision Support System (FDSS) for personalized asthma monitoring, leveraging semantic reasoning

techniques and SPARQL querying to enhance decision-making accuracy and provide individualized assess-

ments of asthma control and exacerbation risk. By utilizing semantic reasoning, the FDSS captures intricate

relationships among asthma parameters, health data, triggers, and treatment outcomes, enabling precise man-

agement decisions. Development involves creating an ontology to encapsulate asthma domain knowledge,

representing fuzzy logic, integrating crisp and fuzzy clinical variables, and executing SPARQL queries for

fuzzy inference. The proposed FDSS demonstrates the feasibility of integrating these techniques for person-

alized asthma management, offering ﬂexibility and adaptability to improve treatment outcomes and quality of

life. Further research is needed to validate its efﬁcacy in real-world healthcare settings.

1 INTRODUCTION

Asthma is a signiﬁcant global health issue, preva-

lent across all ages, particularly affecting children

(WHO, 2019). This chronic lung disease causes air-

way inﬂammation and hyper-responsiveness, lead-

ing to wheezing, breathlessness, chest tightness, and

coughing (Gibson, 2000). Effective asthma manage-

ment requires regular symptom tracking, lung func-

tion assessment, trigger identiﬁcation, and therapeutic

adjustments (Kang, 2024). In 2019, asthma affected

262 million people worldwide, resulting in 455,000

deaths (WHO, 2019). In the U.S., over 25 million

individuals have asthma, including more than 5 mil-

lion children (Cleveland Clinic, 2023). Proper moni-

toring can control symptoms and allow individuals to

lead active lives, while avoiding triggers is crucial for

symptom relief (WHO, 2019). The high asthma mor-

tality in lower-income nations underscores the need

for better diagnostic and treatment strategies. The

World Health Organization aims to reduce the bur-

den of asthma and progress towards universal health

coverage. Asthma is classiﬁed into allergic and non-

allergic types, triggered by factors such as allergens,

air pollution, weather conditions, tobacco smoke, and

https://orcid.org/0000-0002-1051-2814

food allergens (Ajami, 2022). Symptoms vary widely

in frequency and severity, with each individual react-

ing differently. Traditional asthma monitoring meth-

ods often use simplistic decision systems and static

protocols, failing to account for the disease’s com-

plexity. These methods typically categorize patients

as controlled or uncontrolled based on ﬁxed crite-

ria for symptom severity, lung function, or medica-

tion use, which can lead to suboptimal management

(Pinnock, 2015). There is a pressing need for ad-

vanced decision support systems that integrate vari-

ability among patients, environmental factors, and

treatment nuances. A promising approach is combin-

ing logic, semantics, reasoning, rules, and querying

to create a robust framework for personalized asthma

care. Semantic reasoning can capture and interpret

complex relationships between asthma types, symp-

toms, triggers, and treatment outcomes, enhancing

decision-making and management effectiveness.

The integration of Clinical Decision Support Sys-

tems (CDSS) in digital health is crucial for remote

monitoring and decision-making, particularly for

asthma management, leading to improved outcomes

(Dramburg et al., 2020). Key CDSS components

include monitoring, digital technology, decision-

making, and remote communication. Successful

CDSS requires context awareness and personaliza-

Chatterjee, A.

A Fuzzy Decision Support System with Semantic Knowledge Graph for Personalized Asthma Monitoring: A Conceptual Modeling.

DOI: 10.5220/0012910500003838

Paper published under CC license (CC BY-NC-ND 4.0)

In Proceedings of the 16th International Joint Conference on Knowledge Discovery, Knowledge Engineering and Knowledge Management (IC3K 2024) - Volume 2: KEOD, pages 143-150

ISBN: 978-989-758-716-0; ISSN: 2184-3228

143

tion, adapting to individual circumstances and spe-

ciﬁc needs. Al-Dowaihi et al. (Al-Dowaihi, 2013)

developed a prototype for asthma self-management in

high pollution environments, which also alerts health-

care providers. Anantharam et al. (Anantharam,

2015) created kHealth, which uses sensor data to help

physicians determine asthma severity and improve pa-

tient quality of life. Ra et al. (Ra, 2016) introduced

AsthmaGuide, a cloud-based system that uses smart-

phones for real-time data collection and patient em-

powerment. Dieffenderfer et al. (Dieffenderfer, 2016)

developed a wearable sensor system to study the im-

pact of environmental factors on asthma. Quinde et

al. (Quinde, 2018) proposed context-aware systems to

enhance personalized asthma management, while Gy-

rard et al. (Gyrard, 2018) combined multiple knowl-

edge sources to personalize chronic disease manage-

ment. Galante et al. (Galante, 2022) developed

a context-based asthma control method and demon-

strated a self-monitoring approach, though lacking

dynamism. Ajami et al. (Ajami, 2022) created an

ontology-driven model for personalized asthma risk

detection and exacerbation prediction. Vatsal et al.

(Vatsal, 2024) developed an AI ensemble model for

asthma exacerbation forecasting, showing potential

for enhanced prediction accuracy and individualized

treatment. Molﬁno et al. (Molﬁno, 2024) reviewed AI

advancements in asthma management, while Wiec-

zorek et al. (Wieczorek, 2024) evaluated automated

acoustic analysis for asthma diagnosis and monitor-

ing. These studies address heterogeneity in system

modeling, rule-based, or AI-driven decision-making,

but often neglect comprehensive context representa-

tion and uniﬁed knowledge models.

To enhance decision-making accuracy and adapt-

ability, this study introduces a Fuzzy Decision Sup-

port System (FDSS) for personalized asthma moni-

toring, incorporating semantic reasoning techniques.

Fuzzy logic (Wikstr

om, 2014) offers a ﬂexible ap-

proach for handling uncertain and imprecise data,

enabling more precise and context-sensitive deci-

sions. Semantic reasoning, including OWL-based

knowledge representation (Chatterjee, 2021) and

SPARQL querying (Chatterjee, 2022b), further re-

ﬁnes decision-making by capturing intricate data pat-

terns. By integrating fuzzy logic with semantic rea-

soning, this FDSS addresses conventional asthma

monitoring limitations and provides clinicians with

personalized, actionable insights for improved asthma

management. The identiﬁed research questions for

this study are − a. How does a semantic knowledge

graph enhance the development of a FDSS for person-

alized asthma monitoring? b. What are the challenges

and considerations in representing fuzzy knowledge

in FDSS? and c. How does knowledge base, crisp val-

ues and their fuzzy representations, facilitate person-

alized assessments and rule-based decision-making in

asthma management using the FDSS?

This is strictly a technical proof-of-concept study;

rather than a clinical study, and focuses on the con-

ceptual modeling and its theoretical veriﬁcation. The

future study will focus more on technical validation

using robust datasets. The paper is structured as fol-

lows. Section 2 elaborates the proposed FDSS and

associated approaches for system modeling. Section

3 describes the implementation, answers the research

questions, and elaborates study limitations and future

scope. Moreover, the paper is concluded in Section 4.

2 PROPOSED WORK

The design and development of the proposed FDSS

involve following key aspects.

2.1 System Architecture

The proposed FDSS architecture (see Fig. 1) includes

the following layers − Data Acquisition Layer: Col-

lects patient data from sources such as electronic

health records (EHRs), clinical assessments, wearable

devices, and patient-reported data, including context

like asthma attack triggers and factors. Data Prepro-

cessing Layer: Ensures data quality, consistency, and

compatibility through cleaning, normalization, fea-

ture extraction, and transformation. Ontology Con-

struction Layer : Constructs an OWL ontology rep-

resenting asthma monitoring knowledge with classes,

properties, relationships, individuals, logical opera-

tors (AND, OR, NOT), inference rules (TBox and

ABox), and axioms. It integrates relevant ontolo-

gies (Ajami, 2022), such as Asthma from BioPor-

tal, weather from COPDology, food allergens from

FoodOn, and symptoms from SNOMED-CT. Fuzzy

Knowledge Representation Layer: Encodes fuzzy

knowledge in the ontology using linguistic variables,

membership functions, and fuzzy rules, storing both

crisp values and fuzzy representations of clinical vari-

ables. Fuzzy Inference Engine Layer: Uses SPARQL

queries to perform fuzzy inference on patient data for

personalized asthma control and exacerbation risk as-

sessments. Decision Support Layer: Combines fuzzy

inference results with clinical guidelines to provide

actionable recommendations for asthma management.

User Interface Layer: Offers a user-friendly inter-

face with dashboards and visualization tools for clini-

cians to input data, view recommendations, and track

patient progress. Integration Layer: Ensures FDSS

KEOD 2024 - 16th International Conference on Knowledge Engineering and Ontology Development

144

Figure 1: The architecture of the proposed Fuzzy Decision Support System (FDSS).

integration with EHR systems, CDSS, telemedicine

platforms, and Fast Healthcare Interoperability Re-

sources (FHIR), adhering to HL7 standards (Chatter-

jee, 2022a) for interoperability and data exchange.

2.2 Nature of Data

Asthma monitoring involves collecting diverse data

from sensors, questionnaires, and medical surveil-

lance (Merchant, 2018). Personal data includes de-

mographics (age, gender, location, occupation, smok-

ing status), asthma severity (severe, mild, moder-

ate), and physiological metrics (BMI, SpO2, heart

rate, body temperature). Age distinctions are cru-

cial, differentiating between Adult-onset (age < 18)

and Pediatric (age < 5) asthma. Asthma types in-

clude Exercise-induced, Occupational, and Asthma-

COPD overlap syndrome, with causes such as al-

lergies, chemicals, genetics, and infections. Symp-

toms range from intermittent to persistent, includ-

ing chest tightness, coughing, shortness of breath,

and wheezing, with severe attacks showing anxiety,

pain, nasal congestion, and cyanosis. Sensors de-

tect air pollutants like temperature, humidity, and par-

ticulate matter, aiding in trigger identiﬁcation. Sen-

sor data, often continuous, falls into physiological

or environmental categories. Effective asthma man-

agement requires frequent monitoring and personal-

ized decision-making (Chakraborty, 2023), which an

FDSS can support, potentially reducing hospitaliza-

tions and costs. Analyzing both general and sensor

data enables informed treatment adjustments and pre-

ventive measures.

2.3 Ontology and Reasoning

Designing and developing an ontology and reasoning

involves deﬁning formal representations of concepts,

relationships, and inference rules within the ontology.

Proposed ontology can be mathematically represented

as:

O = {C, P, I, φ, R}

where: C = set of classes representing key concepts

in asthma monitoring (e.g., Patient, Symptom, Per-

sonal Information, Medication, TreatmentPlan), P =

set of properties representing relationships between

classes (e.g., hasSymptom, hasMedication, hasTreat-

mentPlan), I = set of individuals or instances of

classes, φ = Set of logical formulas representing ax-

ioms and constraints (e.g., a patient can have multi-

ple symptoms, but only one treatment plan at a time),

and R = set of inference rules for deriving new knowl-

edge from existing data and ontology axioms (e.g., If

a patient has a speciﬁc combination of symptoms, rec-

ommend a personalized treatment plan based on his-

torical data, if a patient’s environmental exposure data

indicates high levels of allergens, suggest adjustments

to the treatment plan to account for potential exacer-

bation). The same can be represented mathematically

as follows − C = {C

, . . . ,C

}, P = {(C

)},

where (C

) denotes a property between classes C

and C

, and

I = {i

, i

, ..., i

where i

is an instance of class C

Φ = {∀x(C(x) → P(x)), ∃x(C(x) ∧ Q(x))}

where C(x) and P(x) are predicates representing class

membership and property relationships, respectively.

A simpliﬁed example using ﬁrst-order logic is as fol-

lows, −

P(x) represent the predicate “x is a patient”.

O(x, y) represent the predicate “x has observation y”.

S(x, y) represent the predicate “x has symptom y”.

T (x, y) represent the predicate “x has treatment plan

y”.

A Fuzzy Decision Support System with Semantic Knowledge Graph for Personalized Asthma Monitoring: A Conceptual Modeling

145

Axioms:

∃p ∈ P : There exists at least one patient in the

ontology

∀x∃y : P(x ) → O(x, y) : Patient has observation plans

∀x∃y : P(x ) → T (x, y) : Patient has a treatment plan

∀x∀y : T (x, y) → R(x, y) : Every treatment plan is

recommended for the patient

Inference Rules:

∀x∀y : P(x ) ∧ S(x,y

) ∧ S(x, y

) ∧ . . .

→ T (x, z) : If a patient exhibits speciﬁc symptoms,

recommend a personalized treatment plan

In description logics (DL), a TBox (T ) represents

the terminology or schema of the ontology, an ABox

represents the assertion or instance data, and an RBox

(R ) represents the relationship between classes and

properties; the same has been incorporated into the

mathematical model for ontology and reasoning in the

context of personalized asthma monitoring. As an ex-

ample, the concepts can be represented as:

T : {∀x(AP(x) → P(x)), x(HS(x, y) → S(x) ∧ P(y))},

A = {AP(a), S(s), HS(a, s)},

where a is an instance of AsthmaPatient class, s is an

instance of Symptom class, and (a, s) asserts the re-

lationship HasSymptom between AsthmaPatient and

Symptom, and

R = {F(HS ), I(HS, SO)},

where F(HS) speciﬁes that the property HasSymp-

tom is functional, and I(HS, SO) indicates the inverse

relationship between HasSymptom and SymptomOf.

Inference rules have been applied to the TBox and

ABox to derive new knowledge from existing asser-

tions and schema deﬁnitions.

Let n be the number of classes, m be the number

of properties, and k be the knowledge base size. The

time complexity for building the ontology model is

O(n + m + k). Consistency checks and completeness

veriﬁcation involve traversing the ontology structure

and logical evaluations. Overall complexity is inﬂu-

enced by the ontology size, inference rules complex-

ity, ontology reﬁnement extent, and test data size.

2.4 Fuzzy Knowledge Representation

Integrating fuzzy knowledge into the personalized

asthma monitoring ontology enhances adaptability

and depth by accommodating uncertainty and vague-

ness in healthcare data. Fuzzy ontology improves

modeling of imprecise concepts, relationships, and

membership degrees, addressing challenges in sub-

jective symptom categorization. Unlike conventional

ontologies that struggle with subjective evaluations

like ”mild,” ”moderate,” or ”severe,” fuzzy ontology

captures these assessments accurately, enabling pre-

cise reasoning and decision-making in asthma care.

Fuzzy sets and membership functions represent lin-

guistic variables and their degrees of membership.

Let’s consider, X: Universe of discourse (set of all

possible values), µ

(x): Membership function repre-

senting the degree of membership of an element x in

the fuzzy set A, and A: Fuzzy set deﬁned over uni-

verse X . Each patient p is associated with fuzzy sets

representing observational variables:

Observations(p) = {(A

, µ

), (A

, µ

), . . . , (A

, µ

)}

where A

represents a linguistic variable and µ

repre-

sents its corresponding membership function. There-

fore, fuzzy logic system (FLS) can be mathematically

represented as:

FLS = (X, {A

, A

, . . . , A

}, {µ

, µ

, . . . , µ

}, Rules)

(x) represents the degree of membership of

of crisp value x in linguistic variable A

The fuzzy sets in the ontology modiﬁes the TBox

(a set of logical formulas deﬁning fuzzy classes, prop-

erties, and relationships), ABox (a set of assertions or

instance data with fuzzy degrees of membership), and

RBox (a set of logical formulas deﬁning fuzzy prop-

erties and their characteristics).

T Box = {∀x(A(x) → µA(x)), ∀x(B(x) → µB(x)),

R(A, B), . . .}

Here, ∀x(A(x) → µA(x)) represents the ﬁrst part of

T Box, ∀x(B(x) → µB(x)) represents the second part,

R(A, B) represents the relation between A and B.

ABox = {(a, µ

(a)), (b, µ

(b)), (c, µ

(c)), . . .}

where µ

(a), µ

(b), µ

(c), etc., represent member-

ship functions.

RBox = {∀x (A(x)∧ B(x) → R

(x)), Functional(R

InverseOf(R

, R

), . . .}

Fuzzy inference rules apply fuzzy logic opera-

tions to fuzzy sets to derive new fuzzy knowledge.

If A is true with degree µ

(x) and B is true

with degree µ

(x), then C is true with degree

min(µ

(x), µ

(x)). As an example, let, Crisp value

Set − Symptom Severity: {Low, Medium, High} ∈

{1, 10}, Medication Adherence: {Low, Medium,

High} ∈ {1, 10}, and Environmental Sensitiv-

ity: {Low, Medium, High} ∈ {1, 10}, Predicate

KEOD 2024 - 16th International Conference on Knowledge Engineering and Ontology Development

146

logic statements − HighSymptomSeverity(x): Pa-

tient x or P(x) has high symptom severity level,

LowMedicationAdherence(x): P(x) has low medi-

cation adherence, HighEnvironmentalSensitivity(x):

P(x) has high environmental sensitivity,

UncontrolledAsthma(x): P(x) has uncontrolled

asthma, and HighExacerbationRisk(x): P(x) has

high exacerbation risk, and Fuzzy rules − Rule −

1: UncontrolledAsthma ← HighSymptomSeverity ∨

LowMedicationAdherence, and Rule − 2: HighEx-

acerbationRisk ← HighEnvironmentalSensitivity.

Let’s fuzzify the predicate logic statements

for Participant P(x) using linguistic variables

and membership functions − Symptom Severity:

High or HighSymptomSeverity(P(x)) (Mem-

bership: 0.8), Medication Adherence: Low or

LowMedicationAdherence(P(x)) (Membership:

0.5), and Environmental Sensitivity: High or

HighEnvironmentalSensitivity(P(x)) (Membership:

0.7). The centroid method (Chakraverty, 2019)

which has been used for defuzziﬁcation, calcu-

lates the center of gravity of the aggregated fuzzy

output to determine the crisp output value for

further decision-making with SPARQL in fuzzy

ontology. According to the fuzzy Rule − 1, the

membership degree of UncontrolledAsthma for

P(x): µ

UncontrolledAsthma

(P(x)) = max(0.8, 0.5) = 0.8,

and According to the fuzzy Rule − 2, the mem-

bership degree of HighExacerbationRisk for P(x):

HighExacerbationRisk

(P(x)) = 0.7. Based on centroid

defuzziﬁcation method, the crisp output for Uncon-

trolledAsthma is 1 and for HighExacerbationRisk is

0.7.

Below are sample queries to retrieve the a. fuzzy

sets for different degrees of symptoms, asthma sever-

ity, and tiredness with their defuzziﬁed membership

values, and b. the current degree of asthma severity

and frequency of symptoms for patients :

Query:1 - Retrieve Fuzzy Sets for Degree of asthma

severity, tiredness, and frequency of symptoms

SELECT ?severity ?tiredness ?frequency ?memvalue

WHERE {

?severity a :DegreeOfAsthmaSeverity ;

?tiredness a :Tiredness ;

?frequency a :FrequencyOfSymptoms ;

:hasMembershipValue ?memvalue .

}

Query:2 - Retrieve Asthma Severity and Symptom

Frequency for Patients

SELECT ?patient ?severity ?symptom ?frequency

WHERE

{

?patient a :Patient ;

:hasAsthmaSeverity ?severity .

?patient :hasSymptom ?symptom .

?symptom :hasFrequency ?frequency .

}

2.5 Proposed Algorithm

Algorithm 1 integrates asthma patient data and fuzzy

ontology for personalized care recommendations. It

evaluates fuzzy rules based on symptoms and demo-

graphics, optimizing care by considering their inter-

action and enhancing recommendation efﬁcacy.

Algorithm 1: Personalized Asthma Monitoring and

Rule-based Decision-Making.

Require: • Patient data including symptoms and

demographic factors

• Fuzzy ontology with properties, linguistic

terms, membership functions, fuzzy rules,

and SPARQL queries.

Ensure: Personalized care recommendations based

on the patient’s symptom proﬁle and demo-

graphic characteristics

1: Retrieve patient observable and measurable data

from Ontology

2: Execute SPARQL Queries to obtain fuzzy out-

comes for each symptom and demographic factor

3: Evaluate fuzzy rules using the fuzzy outcomes

4: Aggregate the fuzzy rule activations

5: Generate personalized care recommendations

based on the aggregated rule activations

6: return Personalized care recommendations

Goal: The algorithm aims to trigger logical rules

of the form (A IMPLIES B) or its equivalent (NOT(A)

OR B) for generating tailored recommendations af-

ter decision-making. If certain variables are inferred

as true, recommendations are provided based on the

originating semantic data. Time Complexity: Re-

trieving patient data has a complexity of O(1) or O(n)

if iterating over ’n’ patients. Executing SPARQL

queries has a complexity of O(m), where ’m’ is the

number of queries. Evaluating fuzzy rules has a com-

plexity of O(r), where ’r’ is the number of rules.

Aggregating fuzzy rule activations and generating

recommendations both have a complexity of O(1).

Therefore, the overall time complexity is O(m + r).

A Fuzzy Decision Support System with Semantic Knowledge Graph for Personalized Asthma Monitoring: A Conceptual Modeling

147

Figure 2: Key classes of the proposed ontology using OWLViz in Prot

Space Complexity: Storing patient data requires

O(n) space. The space required for the fuzzy ontol-

ogy is O(k + r), accounting for linguistic terms and

fuzzy rules. Thus, the overall space complexity is

O(k + r). The time and space complexities primarily

depend on the number of SPARQL queries executed

(m) and the number of fuzzy rules (r). Evaluation:

The concept is evaluated using the “Asthma Disease

Prediction Dataset” (Dataset, 2024), which includes

patient data, environmental factors, and medical his-

tory to predict asthma onset, severity, and treatment

outcomes.

2.6 Experimental Setup

Fuzzy OWL 2 in Prot

e was used to design the on-

tology model, including classes, properties, relation-

ships, and individuals, while the Fuzzy DL reasoner

veriﬁed its consistency.

3 IMPLEMENTATION AND

DISCUSSION

The important classes of fuzzy ontology are depicted

in Fig. 2 (Axiom: 150, Logical axiom: 90, Declara-

tion axiom: 102, Class: 75, object property: 45, data

property: 63, and class axioms: 95). In Prot

e, using

the Fuzzy DL reasoner, it achieved a reasoning time of

under 40.0 seconds without reporting any inconsisten-

cies. When loaded into the Jena workspace with the

“OWL MEM MICRO RULE INF” ontology speciﬁ-

cation in TTL format (OWL full), the reading time

was approximately 4.0-4.5 seconds. Queries for on-

tology classes, ontologies, and statement elements

(predicate, subject, object) using Jena were executed

in under 3.0 seconds, 1.0 seconds, and 4.0 seconds

respectively. Each ontology model, representing a

complete RDF graph, is associated with a document

manager (default global document manager: “Ont-

DocumentManager”) to facilitate ontology document

processing. In the ontology API, all classes repre-

senting ontology values inherit from “OntResource”

with common attributes (versionInfo, comment, la-

bel, seeAlso, isDeﬁnedBy, sameAs, differentFrom)

and methods (add, set, list, get, has, remove). This

paper presents a mathematical model for personal-

ized recommendation generation based on SPARQL

queries on top of a fuzzy ontology. The model uti-

lizes set theory to deﬁne patient attributes, ﬁlter func-

tions, and recommendation functions. It dynamically

generates tailored recommendations for asthma man-

agement based on patient characteristics and severity

of their asthma condition. The used dataset helps in

this regard for the proposed theoretical concept evalu-

ation. Let, P represent the set of patients in the fuzzy

ontology. Each patient p ∈ P has attributes such as

tiredness T (p), difﬁculty in breathing D(p), severity

S(p), and age A(p). Deﬁned ﬁlter function F to select

patients, based on speciﬁc criteria:

F(p) =

(

1 if T (p) ≥ t

min

and D(p) ≥ d

min

and A(p) ≥ a

min

0 otherwise

Where:

min

is the minimum threshold for tiredness,

min

is the minimum threshold for difﬁculty

in breathing,

min

is the minimum age threshold.

Next, let R denote the set of recommendations.

For each patient p selected by F, the generated

personalized recommendations r are based on their

severity S(p) using fuzzy logic rules. The recommen-

dation function G is deﬁned as:

G(p) =



“Administer emergency treatment” if S(p) = “Severe”

“Increase medication dosage” otherwise

KEOD 2024 - 16th International Conference on Knowledge Engineering and Ontology Development

148

Finally, a SPARQL query has been executed on the

fuzzy ontology to ﬁlter patients based on F, and for

each selected patient, generate personalized recom-

mendations using G. The result is a set of patient-

recommendation pairs {(p

, r

), (p

, r

), . . .}, provid-

ing tailored asthma management recommendations

for each patient.

The proposed fuzzy ontology includes properties,

such as “Age Group”, “Degree of Asthma Sever-

ity”, “Frequency of Symptoms”, “Effectiveness of

Medication”, and “Recommendation”. Fuzzy rules

recommend treatment adjustments based on these

properties. Moreover, speciﬁc symptoms and de-

mographic factors are incorporated into the ontol-

ogy to personalize asthma monitoring and care. The

logical structure of the ontology is supported by

the features of the adopted dataset. The extended

structure of the fuzzy ontology includes properties

such as “Tiredness” (Tiredness(x), where x represents

the degree of tiredness experienced by the patient,

and membership functions: µ

Low

(x), µ

Medium

(x),

and µ

High

(x)), “Dry-Cough” (Dry-Cough, where

Dry-Cough(x) is true if the patient experiences a dry

cough and false otherwise), “Difﬁculty-in-Breathing”

(Difﬁculty-in-Breathing(x), where x represents the

severity of difﬁculty in breathing, and membership

functions: µ

Low

(x), µ

Medium

(x), µ

High

(x)), “Sore-

Throat” (Sore-Throat, where Sore-Throat(x) is true if

the patient experiences a sore throat and false other-

wise), “Severity” (Severity(x), where x represents the

severity levels, and membership function: µ

Mild

(x),

Moderate

(x), µ

Severe

(x)), and Age with corresponding

linguistic terms and membership functions. Age has

been found as an important property in asthma moni-

toring and care. Thus, let, A

, A

, and A

represent

the fuzzy sets for the following age categories: A

(x):

Less than 5 years, A

(x): 5-18 years, A

(x): 18-65,

and A

(x): Greater than 65 years.

For evaluation on the adopted dataset, two fuzzy

rules have been considered from the rule-base or

knowledge base (KB) for personalized recommen-

dation generation: Rule − 3: IF Tiredness is high

AND DifﬁcultyInBreathing is moderate THEN In-

crease Medication Dosage (IMD), and Rule − 4:

IF Severity is severe THEN Administer Emergency

Treatment (AET). The defuzziﬁed decision is ob-

tained through the centroid method after applying

fuzzy rules and obtaining fuzzy outputs.

An example SPARQL query for generating per-

sonalized asthma recommendations based on speciﬁc

criteria and fuzzy properties is:

SELECT ?recommendation

WHERE {

?patient :hasSymptom ?symptom .

?symptom :Tiredness ?tiredness .

?symptom :Difﬁculty-in-Breathing ?breathing .

?symptom :Severity ?severity .

?patient :hasDemograph ?demo .

?demo :hasAge ?age .

FILTER (?tiredness ≥ 0.7 && ?breathing ≥ 0.5

&& ?age ≥ 18)

BIND (IF(?severity = :Severe, “AET”,

“IMD”) AS ?recommendation)

}

Patient data, including tiredness, breathing difﬁ-

culty, and severity, is ﬁltered to identify adults with

high tiredness and moderate breathing difﬁculty. Rec-

ommendations for increasing medication or adminis-

tering emergency treatment are made based on fuzzy

rules and severity. The aggregated fuzzy output is

defuzziﬁed using the centroid method to provide a

clear recommendation. For example, if a patient has

tiredness = High (0.8), DifﬁcultyInBreathing = Mod-

erate (0.6), and Severity = Moderate (0.4), the sys-

tem generates a recommendation to ”Increase Med-

ication Dosage” based on Rule 3. If the centroid

value exceeds a threshold (e.g., 0.5), the recommen-

dation is conﬁrmed. The implementation addresses

the research questions as follows: First, the seman-

tic knowledge graph, shown in Fig. 2, underpins the

FDSS for personalized asthma monitoring, efﬁciently

handling reasoning tasks with quick processing in

Prot

e and Jena environments. Second, although

representing fuzzy knowledge in the FDSS is com-

plex, the ontology integrates fuzzy rules effectively,

enabling tailored recommendations based on proper-

ties like tiredness and breathing difﬁculty. Third, the

synergy between the knowledge base, crisp values,

and their fuzzy representations supports personalized

monitoring and decision-making. SPARQL queries

use criteria such as tiredness and breathing difﬁculty

to generate tailored recommendations, with defuzzi-

ﬁcation providing clear insights for asthma manage-

ment. Overall, the algorithm integrates patient data,

rule-based logic, and fuzzy ontology via SPARQL

queries to deliver customized asthma monitoring rec-

ommendations. In this context, fuzzy ontology sur-

passes general ontology by effectively capturing and

representing uncertainty and imprecision in patient

data, thus enhancing decision-making accuracy. It im-

proves personalized asthma monitoring by handling

uncertainty, offering ﬂexibility, and integrating with

A Fuzzy Decision Support System with Semantic Knowledge Graph for Personalized Asthma Monitoring: A Conceptual Modeling

149

semantic web technologies. Fuzzy ontology-based

SPARQL enhances interpretability and adaptability,

though scalability issues may arise with growing data

volumes. Future research could integrate machine

learning models to address these scalability concerns

and improve predictive power and automatic feature

learning in the FDSS for asthma management.

4 CONCLUSION

The FDSS for personalized asthma monitoring lever-

ages semantic reasoning and fuzzy logic to enhance

asthma care. By using ontological representation,

fuzzy reasoning, and SPARQL queries, the FDSS of-

fers a ﬂexible framework for personalized decision-

making, potentially improving treatment effective-

ness and patient quality of life. Future work should

focus on validating the FDSS’s effectiveness and

practicality in real-world healthcare settings.

REFERENCES

Ajami, H. (2022). First steps of asthma management

with a personalized ontology model. Future Internet,

14(7):190.

Al-Dowaihi, D. (2013). Mbreath: Asthma monitoring sys-

tem on the go. In 2013 International Conference on

Computer Medical Applications (ICCMA), pages 1–4.

IEEE.

Anantharam, P. (2015). Knowledge-driven personalized

contextual mhealth service for asthma management in

children. In 2015 IEEE international conference on

mobile services, pages 284–291. IEEE.

Chakraborty, A. (2023). Development of an iot-enabled

cost-effective asthma patient monitoring system: In-

tegrating health and indoor environment data with sta-

tistical analysis and data visualization. Internet of

Things, 24:100942.

Chakraverty, S. (2019). Defuzziﬁcation. Concepts of

Soft Computing: Fuzzy and ANN with Programming,

pages 117–127.

Chatterjee, A. (2021). An automatic ontology-based ap-

proach to support logical representation of observ-

able and measurable data for healthy lifestyle man-

agement: Proof-of-concept study. Journal of Medical

Internet Research, 23(4):e24656.

Chatterjee, A. (2022a). Hl7 fhir with snomed-ct to achieve

semantic and structural interoperability in personal

health data: a proof-of-concept study. Sensors,

22(10):3756.

Chatterjee, A. (2022b). Personalized recommendations

for physical activity e-coaching (ontorecomodel): on-

tological modeling. JMIR Medical Informatics,

10(6):e33847.

Cleveland Clinic (2023).

https://my.clevelandclinic.org/health/diseases/6424-

asthma.

Dataset (2024). https://www.kaggle.com/datasets/

deepayanthakur/asthma-disease-prediction.

Dieffenderfer, J. (2016). Low-power wearable systems for

continuous monitoring of environment and health for

chronic respiratory disease. IEEE journal of biomedi-

cal and health informatics, 20(5):1251–1264.

Dramburg, S., Marchante Fern

andez, M., Potapova, E., and

Matricardi, P. M. (2020). The potential of clinical de-

cision support systems for prevention, diagnosis, and

monitoring of allergic diseases. Frontiers in Immunol-

ogy, 11:575892.

Galante, C. M. (2022). Asthma management updates. Nurs-

ing2023, 52(2):25–34.

Gibson, P. G. (2000). Monitoring the patient with asthma:

an evidence-based approach. Journal of Allergy and

Clinical Immunology, 106(1):17–26.

Gyrard, A. (2018). Personalized health knowledge graph.

In CEUR workshop proceedings, volume 2317. NIH

Public Access.

Kang, H.-q. (2024). The experience of shared decision-

making for people with asthma: A systematic review

and metasynthesis of qualitative studies. Health Ex-

pectations, 27(2):e14039.

Merchant, R. (2018). Digital health intervention for asthma:

patient-reported value and usability. JMIR mHealth

and uHealth, 6(6):e7362.

Molﬁno, N. A. (2024). Machine learning approaches to pre-

dict asthma exacerbations: A narrative review. Ad-

vances in Therapy, 41(2):534–552.

Pinnock, H. (2015). Supported self-management for

asthma. Breathe, 11(2):98–109.

Quinde, M. (2018). Personalisation of context-aware so-

lutions supporting asthma management. In Comput-

ers Helping People with Special Needs: 16th Interna-

tional Conference, ICCHP 2018, Linz, Austria, July

11-13, 2018, Proceedings, Part II 16, pages 510–519.

Springer.

Ra, H.-K. (2016). Asthmaguide: an asthma monitoring

and advice ecosystem. In 2016 IEEE Wireless Health

(WH), pages 1–8. IEEE.

Vatsal, S. K. (2024). Advanced ensemble learning approach

for asthma prediction: Optimization and evaluation.

In 2024 International Conference on Automation and

Computation (AUTOCOM), pages 283–288. IEEE.

WHO (2019). https://www.who.int/news-room/fact-

sheets/detail/asthma.

Wieczorek, K. (2024). Acoustic biomarkers in asthma:

a systematic review. Journal of Asthma, (just-

accepted):1–18.

Wikstr

om, R. (2014). Fuzzy ontology for knowledge mo-

bilisation and decision support. PhD thesis,

Abo

Akademi-

Abo Akademi University.

KEOD 2024 - 16th International Conference on Knowledge Engineering and Ontology Development

150