Analysis of Health Indicators for Heart Disease Based on Formal
Concept Analysis
Laura Xavier
a
, Julio Neves
b
, Luiz Zarate
c
and Mark Song
d
Universidade Cat
´
olica de Minas Gerais, Computer Science Department, Minas Gerais, Brazil
Keywords:
Formal Concept Analysis, FCA, Health Indicators, Heart Disease, Lattice Miner.
Abstract:
This study addresses the global concern of cardiovascular health by analyzing key risk factors such as high
blood pressure, cholesterol levels, and smoking habits, which contribute to the onset of heart disease. Using
Formal Concept Analysis (FCA), a mathematical framework for uncovering relationships in complex datasets,
this research examines a health dataset of over 200,000 records to identify critical behavioral and health in-
dicators related to cardiovascular problems. Although 80 association rules were extracted, 12 were selected
for detailed analysis due to their significance in both risk and protective factors. Key findings reveal strong
correlations between physical inactivity, poor dietary habits, and the likelihood of heart disease, providing ac-
tionable insights for healthcare professionals and policymakers. This study aims to deepen the understanding
of cardiovascular risk factors and support the development of more effective prevention measures to improve
global health outcomes.
1 INTRODUCTION
In the field of computer science and data analy-
sis, advanced analytical methods have become essen-
tial for extracting meaningful insights from complex
datasets. One such method is Formal Concept Analy-
sis (FCA), a mathematical approach rooted in set the-
ory and logic. FCA provides a structured framework
for organizing data, uncovering patterns and relation-
ships that might remain hidden in conventional anal-
yses.
This study applies FCA to the public “Heart Dis-
ease Health Indicators Dataset,” available on Kaggle
(Teboul, 2022). This dataset, which includes over
200,000 records of patient health data, offers a valu-
able opportunity to explore the relationships between
various health indicators and the likelihood of heart
disease. By mapping key dimensions such as Body
Mass Index (BMI), physical activity, cholesterol lev-
els, and diabetes, this analysis aims to uncover risk
factors and interdependencies that contribute to heart
disease.
Cardiovascular diseases (CVDs) are the leading
a
https://orcid.org/0009-0008-7751-9640
b
https://orcid.org/0000-0002-0520-9976
c
https://orcid.org/0000-0002-3018-888X
d
https://orcid.org/0000-0001-5053-5490
cause of death worldwide, responsible for an es-
timated 17.9 million deaths annually, according to
the World Health Organization (Organization, 2021).
CVDs encompass a range of disorders affecting the
heart and blood vessels, such as coronary heart dis-
ease and cerebrovascular disease. Many of these con-
ditions can be prevented or mitigated by addressing
behavioral risk factors, such as tobacco use, unhealthy
diets, physical inactivity, and excessive alcohol con-
sumption, all of which are captured in the dataset.
Through this analysis, we aim to provide infor-
mation that can support healthcare professionals and
policymakers in designing more effective preventive
measures and treatment strategies for reducing the im-
pact of cardiovascular diseases.
2 RELATED WORK
According to the paper Formal Concept Analysis –
Overview and Applications (
ˇ
Skopljanac Ma
ˇ
cina and
Bla
ˇ
skovi
´
c, 2014), FCA is a method for knowledge
representation, information management, and data
analysis. It identifies and visualizes all concepts and
their dependencies from tabular input data. The re-
sulting structure of concepts is hierarchically orga-
nized into a concept lattice, which can be presented
as a Hasse diagram. The method is based on ap-
540
Xavier, L., Neves, J., Zarate, L. and Song, M.
Analysis of Health Indicators for Heart Disease Based on Formal Concept Analysis.
DOI: 10.5220/0013158100003911
Paper published under CC license (CC BY-NC-ND 4.0)
In Proceedings of the 18th International Joint Conference on Biomedical Engineering Systems and Technologies (BIOSTEC 2025) - Volume 2: HEALTHINF, pages 540-547
ISBN: 978-989-758-731-3; ISSN: 2184-4305
Proceedings Copyright © 2025 by SCITEPRESS – Science and Technology Publications, Lda.

plied lattice theory and set theory, with applications in
fields such as mathematics, medicine, biology, soci-
ology, psychology, economics, and particularly com-
puter science.
The study highlights FCA as a valuable tool for
designing exams and analyzing results. Hierarchi-
cally organized concepts can be used to determine the
optimal order of exam questions, thus facilitating stu-
dent comprehension and evaluation. An example is
given with an electrical engineering exam, showing
how this approach can provide meaningful insights
into student performance and question difficulty.
In summary, the literature demonstrates how FCA
can be effectively applied to derive conclusions from
datasets by providing valuable insights into the re-
lationships between attributes and elements, leading
to potential improvements and problem-solving ap-
proaches (Ananias et al., 2021) (Miranda et al., 2024).
In another study, (Song et al., 2024) explored the
application of FCA in a triadic approach to charac-
terize infant mortality in different regions of Minas
Gerais, Brazil. Determinant factors such as birth
weight, gestation, and APGAR scores were identi-
fied. The findings revealed associations between vari-
ous variables, highlighting the importance of maternal
education and prenatal care consultations.
3 BACKGROUND
This section aims to review the main tools and
methodologies employed in this study, including For-
mal Concept Analysis and the Lattice Miner software
3.1 Cardiovascular Diseases
Cardiovascular diseases (CVDs) involve a broad
spectrum of disorders that affect the heart and blood
vessels, and their impact on global health is well-
documented. Among the most prevalent conditions
are coronary artery disease (CAD), which results
from the buildup of plaques in the arteries (atheroscle-
rosis), leading to reduced blood flow to the heart, and
cerebrovascular diseases, such as stroke, where the
brain’s blood supply is disrupted (James and Smith,
2023). Both conditions share common risk factors, in-
cluding high blood pressure, high cholesterol, smok-
ing, and poor lifestyle choices.
The link between these modifiable risk factors and
the development of CVDs has been a significant focus
in public health. Regular physical activity, a healthy
diet, and controlling blood pressure and cholesterol
are crucial strategies for preventing these diseases
(Lancet, 2020). While the impact of CVDs is well-
known, there remains a need for ongoing research into
the specific patterns and behaviors that increase the
likelihood of these conditions. In recent years, ad-
vances in data analysis have allowed for a more nu-
anced understanding of how these risk factors inter-
act, particularly through the use of methods like For-
mal Concept Analysis (FCA).
3.2 Formal Concept Analysis
Formal Concept Analysis (FCA) is a mathematical
framework originally developed by Rudolf Wille in
the 1980s for the formal representation of conceptual
knowledge (Wille, 1982).
The method structures data into what is known
as a formal context, which consists of three compo-
nents: a set of objects, a set of attributes, and the in-
cidence relation that links objects to their attributes.
This formal context can be represented as a triple
K = (G, M, I), where:
• G represents the set of objects in the dataset,
• M denotes the set of attributes,
• I ⊆ G ×M is the incidence relation, indicating the
association between objects and their attributes,
where (g, m) ∈ I means that object g has attribute
m.
Given a subset of objects A ⊆ G, the correspond-
ing set of attributes shared by all objects in A is repre-
sented as:
A
′
:=
{
m ∈ M | ∀g ∈ A : (g, m) ∈ I
}
.
Similarly, for a subset of attributes B ⊆ M, the set of
objects that possess all attributes in B is:
B
′
:=
{
g ∈ G | ∀m ∈ B : (g, m) ∈ I
}
.
A formal concept is then defined as a pair (A, B),
where:
• A (the extent) is the set of objects that share the
attributes in B,
• B (the intent) is the set of attributes shared by all
objects in A.
For (A, B) to be a valid formal concept, it must satisfy
A = B
′
and B = A
′
. The collection of all such concepts
forms a concept lattice, a structure that organizes the
formal concepts in a hierarchical order based on their
extents and intents (Ganter and Wille, 2012). This
lattice is denoted as β(K) and provides a visual repre-
sentation of the relationships among concepts.
FCA has a natural application in extracting associ-
ation rules from the formal context. These rules take
the form A → B, meaning that if an object has the at-
tributes in A, it is likely to also have the attributes in B.
The quality of an association rule is evaluated through
two key metrics:
Analysis of Health Indicators for Heart Disease Based on Formal Concept Analysis
541

• Support (s): This measures the proportion of ob-
jects that contain both the attributes in A and B,
calculated as:
s =
|
A
′
∩ B
′
|
|
G
|
,
which reflects the frequency of the rule in the
dataset.
• Confidence (c): This indicates how often the at-
tributes in B appear in the objects that already
have the attributes in A, defined as:
c =
|
A
′
∩ B
′
|
|
A
′
|
.
A confidence of 100% means the rule is an impli-
cation, implying that all objects with the attributes
in A also have those in B.
Through this approach, FCA allows for the dis-
covery of patterns that might not be evident in tradi-
tional data analysis methods. For instance, it can be
used to group patients with similar health indicators
or identify the relationships between risk factors and
diseases, as is the case in this study. The resulting
concept lattice and association rules provide an inter-
pretable and hierarchical view of the data, which is
especially valuable in fields like medicine, where un-
derstanding complex relationships is critical (Stumme
et al., 2002).
3.3 Lattice Miner
Lattice Miner 2.0 is a tool developed at the LARIM
laboratory at the Universit
´
e du Qu
´
ebec en Outaouais,
designed to implement Formal Concept Analysis
(FCA). Through this tool, formal concepts can be
extracted, visualized, and explored from binarized
datasets, organizing them into a lattice structure that
reveals underlying patterns and relationships among
the analyzed attributes (LARIM, 2017).
In this study, Lattice Miner will be used to facili-
tate the discovery of patterns among health indicators,
applying FCA to explore the relationships between
risk factors for heart disease.
4 METHODOLOGY
This section outlines the dataset used in the analy-
sis of heart disease health indicators, detailing the
data preparation process and the application of For-
mal Concept Analysis (FCA) through the use of the
Lattice Miner software.
4.1 Heart Disease and Health Indicators
Dataset
The source of information used in this study is the
Heart Disease Health Indicators Dataset, publicly
available on Kaggle (Teboul, 2022).
This dataset is derived from a larger record
from the Centers for Disease Control and Prevention
(CDC) and contains data from the Behavioral Risk
Factor Surveillance System (BRFSS) survey con-
ducted in 2015 (Dane and Centers for Disease Control
and Prevention, 2015).
The survey includes health-related risk behaviors,
chronic health conditions, and the use of preventive
health services, containing over 400,000 responses
and 330 features.
Alex Teboul, author of the Kaggle record, per-
formed significant preprocessing steps to clean the
data and reduce its dimensionality, as detailed in the
dataset’s accompanying notebook (Teboul, 2022).
The process included maintaining only the most
relevant attributes for predicting heart disease, such
as Body Mass Index (BMI), age, gender, physical and
mental health, smoking status and alcohol consump-
tion.
The target variable is a binary attribute indicating
whether the respondent has ever been diagnosed with
coronary heart disease (CHD) or myocardial infarc-
tion (heart attack).
These efforts resulted in a refined dataset, which
contemplate 253,680 responses and 22 health indica-
tors (Table 1), creating the basis for the analysis in
this research.
4.2 Data Processing
In this study, additional steps were taken to tailor
the dataset further. Specifically, socioeconomic vari-
ables were excluded because, while they are impor-
tant in broader public health analyses, they are not di-
rectly relevant to the objective of this research, which
focuses on health and behavioral indicators associ-
ated with heart disease. The following attributes
were removed from the dataset: Income, Education,
NoDocbcCost, and AnyHealthcare.
After excluding these variables, the dataset was
split into two groups:
• Group 1: Individuals who had experienced heart
disease or a heart attack.
• Group 2: Individuals who had not experienced
heart disease or a heart attack.
This separation allowed for a more targeted analysis
of the factors that contribute to heart disease within
HEALTHINF 2025 - 18th International Conference on Health Informatics
542

Table 1: Attributes and Their Information.
Column Name Description Values
HighBP Has high blood pressure 0 - 1
HighChol Has high cholesterol 0 - 1
CholCheck
Cholesterol checked
0 - 1
within past five years
BMI Body Mass Index 12 - 98
Smoker Smoking status 0 - 1
Stroke Has had a stroke 0 - 1
Diabetes Diabetes diagnosis 0 - 2
PhysActivity Does physical activity 0 - 1
Fruits
Regular fruit
0 - 1
consumption
Veggies
Regular vegetable
0 - 1
consumption
HvyAlcoholCon
Heavy alcohol
0 - 1
consumption
AnyHealthcare Access to healthcare 0 - 1
NoDocbcCost
Could not see
0 - 1
doctor due to cost
GenHlth General health status 1 - 5
MentHlth Mental health status 0 - 30
PhysHlth Physical health status 0 - 30
DiffWalk Difficulty walking 0 - 1
Sex Gender 0 - 1
Age Age range 1 - 13
Education Education level 1 - 6
Income Income level 1 - 8
More details about the possible values of each attribute
can be found in the original
author’s notebook on Kaggle. (Teboul, 2022)
each population.
Subsequently, a series of custom Python algo-
rithms were applied to binarize the dataset, transform-
ing the attributes into binary labels to facilitate anal-
ysis and make the data compatible with the Lattice
Miner software.
In this process, each attribute was categorized into
distinct health indicators (Table 2). For instance,
Body Mass Index (BMI) was categorized as “Over-
weight” for individuals with a BMI over 24, while
those within a normal range were left blank.
This threshold was selected based on the World
Health Organization (WHO) guidelines, which clas-
sify a BMI between 25 and 29.9 as “Overweight” (Or-
ganization, 2000). Age was split into “Young,” “Mid-
dleAged,” and “Senior” categories.
Other health-related attributes, such as fruit and
vegetable consumption, mental and physical health,
and diabetes status, were similarly transformed. The
target variable was also binarized, with one column
indicating whether an individual had a heart problem
or not.
At the conclusion of this process, both groups
were left with 19 attributes each, the only difference
being the target variable.
In Group 1, the target variable HasHeartProb was
set to true for all individuals, indicating they had heart
disease or experienced a heart attack.
Conversely, in Group 2, the target variable No-
HeartProb was used, signifying that these individu-
als had not experienced any heart problems. Group
1 contains 23,893 instances, while Group 2 contains
229,787 instances.
This binarization process simplified the dataset
and allowed for the identification of key health behav-
iors and conditions that correlate with heart disease.
For instance, poor mental or physical health, obesity,
and diabetes were labeled as “risk factors,” while in-
dividuals who showed healthy behaviors, such as reg-
ular consumption of fruits and vegetables, were also
identified.
4.3 Using Lattice Miner
Finally, the binarized data was processed using Lat-
tice Miner (LARIM, 2017), to apply Formal Concept
Analysis. This tool was used to extract association
rules and generate concept lattices, providing a struc-
tured way to identify patterns and relationships be-
tween health behaviors and the occurrence of heart
disease. The resulting rules can evidence how dif-
ferent combinations of factors may contribute to heart
disease, forming the basis for the discussions and con-
clusions in this study.
5 RESULTS AND ANALYSIS
After applying the method, approximately 80 rules
were extracted from the dataset. One example is
CholCheck and PhysActivity ⇒ NoHeartProb. This
rule states that individuals who check their choles-
terol and engage in physical activities have no heart
problems, with a support of 73% and a confidence of
100%.
From the extracted rules, we selected 12 for de-
tailed analysis in this work, covering both groups un-
der study.
5.1 Group 1: Individuals with Heart
Problems
The analysis of Group 1, which consists of individ-
uals with heart problems, reveals several important
patterns regarding their health-related behaviors. This
section will delve into the rules extracted and discuss
Analysis of Health Indicators for Heart Disease Based on Formal Concept Analysis
543

Table 2: New Attributes and Their Information After Data Processing.
Original Attribute Binarized Attribute Precessing Description
BMI Overweight
Marked if individual has
BMI over 24
Age
Young, MiddleAged Categorized by ranges 1 - 3,
Senior 5 - 9 and 10 - 13
Fruits
Fruits/Veggies
Has regular fruit and/or
Veggies
vegetable consumption
MentHlth BadMentHlth
Marked if individual has
MentHlth over 15
PhysHlth BadPhysHlth
Marked if individual has
PhysHlth over 15
GenHlth BadGenHlth
Marked if individual has
GenHlth over 3
Diabetes PD/Diabetes
Has diabetes, pre-diabetes
or borderline diabetes
HeartDiseaseorAttack
HasHeartProb Categorized by whether the
NoHeartProb patient has or not a heart problem
The remaining attributes from the original dataset were retained without modification.
their implications in detail. The rules will be refer-
enced by their indices, as specified in Table 3.
Table 3: Association Rules for Group 1 (Individuals with
Heart Problems).
Rule Supp. Conf.
1 HasHeartProb → CholCheck 98% 98%
2 CholCheck → HasHeartProb 98% 100%
3 HasHeartProb → HighBP 84% 85%
4 HighBP → HasHeartProb 84% 100%
5 HasHeartProb → Male 57% 57%
5.1.1 Cholesterol Check and Heart Problems
One of the most significant rules extracted from
Group 1 is the Rule 1. This rule indicates that 98%
of the individuals with heart problems in this group
regularly undergo cholesterol checks. The confidence
value suggests a very strong relationship, meaning
that having heart problems is highly predictive of en-
gaging in this health behavior.
This rule aligns with what we would expect from
individuals who are already diagnosed with heart con-
ditions. Regular monitoring of cholesterol is crucial
for managing cardiovascular health, as high choles-
terol is a well-known risk factor that can exacerbate
heart conditions. A logical conclusion is that individ-
uals in this group likely undergo routine cholesterol
checks as part of their ongoing medical care to prevent
further complications, such as stroke or heart attack.
The high support and confidence of this rule un-
derscore the importance of cholesterol monitoring in
the medical management of heart disease. It reflects
how medical interventions and self-care practices are
intertwined in the lives of those with heart problems.
A complementary rule, Rule 2, was also extracted
with a similar level of strength. This rule has a support
of 0.98 and a confidence of 1.0, indicating that every
individual who had their cholesterol checked within
the past 5 years in this group has heart problems.
This might initially appear surprising, as we might ex-
pect some individuals without heart problems to also
check their cholesterol.
However, this result is reflective of the group be-
ing analyzed, which specifically includes individuals
with heart problems. The fact that every individual
who checks their cholesterol in this group has heart
problems is simply a consequence of the way the
groups were defined. It is important to note that this
rule applies to Group 1 and does not necessarily imply
that cholesterol checks are exclusively performed by
individuals with heart problems in the general pop-
ulation. Rather, it highlights that within this spe-
cific group, cholesterol monitoring is a near-universal
practice among those managing heart conditions.
5.1.2 High Blood Pressure and Heart Problems
Another key rule extracted from Group 1 is Rule 3,
which indicates that 84% of individuals with heart
problems also suffer from high blood pressure, with
a high confidence of 85%. This suggests a strong as-
sociation between these two attributes.
The connection between heart disease and high
blood pressure is well-established in medical litera-
ture. High blood pressure, or hypertension, is a major
HEALTHINF 2025 - 18th International Conference on Health Informatics
544

risk factor for heart disease, and it is common for in-
dividuals with one condition to also have the other.
The high support of this rule reflects the prevalence
of this comorbidity in Group 1, indicating that man-
aging blood pressure is a critical component of care
for individuals with heart problems.
The confidence level further strengthens the va-
lidity of this rule, implying that it is very likely for
someone with heart problems to also have high blood
pressure. This finding emphasizes the need for inte-
grated care approaches that address both conditions
simultaneously.
Additionally, Rule 4, indicates that every individ-
ual with high blood pressure in Group 1 also has heart
problems. Similar to the rule regarding cholesterol
checks, this result is reflective of the group under
analysis. Since this group consists of individuals with
heart problems, it is not surprising that all individuals
with high blood pressure also have heart issues.
The perfect confidence level suggests that, within
this group, high blood pressure is not seen in isolation
but rather in conjunction with heart problems.
5.2 Group 2: Individuals without Heart
Problems
The analysis of Group 2, which contains individuals
without heart problems, provides a complementary
perspective on health behaviors. This section anal-
ysed the rules extracted from Group 2, which will also
be referenced by their indices, as specified in Table 4.
5.2.1 Cholesterol Check and No Heart Problems
One of the primary rules extracted from Group 2 is
Rule 6. This rule indicates that 95% of individu-
als without heart problems in this group still perform
cholesterol checks, with a high confidence level of
95%.
Table 4: Association Rules for Group 2 (Individuals without
Heart Problems).
Rule Supp. Conf.
6 NoHeartProb → CholCheck 95% 95%
7 CholCheck → NoHeartProb 95% 100%
8 NoHeartProb → PhysActivity 76% 76%
9 PhysActivity → NoHeartProb 76% 100%
10 NoHeartProb → Overweight 70% 70%
11 Overweight → PhysActivity 52% 74%
12 NoHeartProb → Female 57% 57%
It may seem to contrast with the similar rule in
Group 1, where cholesterol checks were strongly as-
sociated with the presence of heart problems. How-
ever, in Group 2, this behavior likely reflects preven-
tive health practices. Individuals in this group may
undergo cholesterol checks as a way to monitor and
manage their cardiovascular risk factors, even though
they have not yet developed heart problems.
The high prevalence of cholesterol checks in this
group also underscores the importance of regular
monitoring in maintaining good health and prevent-
ing the onset of cardiovascular diseases. It suggests
that individuals without heart problems are proac-
tive about their health, using cholesterol checks as a
way to detect potential issues before they develop into
more serious conditions.
The reverse rule, Rule 7, was also a highlight. It
indicates that every individual who have their choles-
terol checked in this group does not have heart prob-
lems. While similar in structure to the corresponding
rule in Group 1, the context is different.
In Group 2, this rule reflects the fact that choles-
terol checks are being performed by individuals as
part of their preventive health measures. Since this
group specifically excludes individuals with heart
problems, it is expected that everyone who keeps
cholesterol monitoring in this group does not have
heart problems. The perfect confidence level supports
this conclusion.
It is possible to conclude that regular cholesterol
monitoring is not limited to those with existing health
issues but is also a common practice among those
without heart problems. It reinforces the notion that
cholesterol checks are an important component of
preventive care in maintaining cardiovascular health.
5.2.2 Physical Activity and No Heart Problems
Physical activity is another important factor. The Rule
8, reveals that 76% of individuals without heart prob-
lems engage in physical activity, reiterating the health
benefits of exercising.
Physical activity is widely recognized as a key fac-
tor for preventing heart disease and maintaining over-
all health. The support for this rule indicates that the
majority of individuals in Group 2 engage in regular
physical activity, likely as part of their efforts to stay
healthy and avoid the development of cardiovascular
problems. The confidence level further suggests that
physical activity is a reliable indicator of the absence
of heart problems in this group.
The complementary Rule 9, suggests that all indi-
viduals who engage in physical activity in Group 2 do
not have heart problems.
The perfect confidence level supports the idea that
regular physical activity is highly effective in prevent-
ing the development of heart problems. This finding
aligns with the extensive body of research that shows
Analysis of Health Indicators for Heart Disease Based on Formal Concept Analysis
545

the protective effects of physical exercise on cardio-
vascular health.
5.2.3 Overweight and No Heart Problems
Another interesting rule from Group 2 is Rule 10,
which suggests that 70% of individuals without heart
problems in this group are overweight. This result
may seem counterintuitive, as being overweight is of-
ten associated with a higher risk of heart disease.
However, this rule highlights that being over-
weight does not automatically lead to heart prob-
lems, particularly if individuals are engaging in other
healthy behaviors, such as regular physical activity.
In fact, another notable rule is Rule 11, which indi-
cates that 52% of individuals in Group 2 who are over-
weight also engage in regular physical activity, and
there’s a 74% likelihood that someone who is over-
weight will also be physically active.
Interestingly, when analyzing Group 1 (individu-
als with heart problems), no relevant rules were found
linking being overweight with physical activity. This
suggests that individuals with heart problems who are
overweight may be less likely to engage in regular ex-
ercise compared to those in Group 2, potentially indi-
cating a gap in their health management strategies.
Together, these two rules suggest that while a ma-
jority of individuals without heart problems may be
overweight, many of them are proactively managing
their health through preventive behaviors, like physi-
cal activity, which can significantly reduce their risk
of developing cardiovascular issues in the future. This
highlights the nuanced relationship between weight
and heart health: although excess weight is often con-
sidered a risk factor, it is not necessarily determinis-
tic of poor cardiovascular outcomes, especially when
healthy behaviors are in place.
5.3 Gender and No Heart Problems
In Group 2, the Rule 12 reveals that 57% of individu-
als without heart problems are female. This suggests
that women in this group are less prone to heart is-
sues, which aligns with broader epidemiological find-
ings indicating that women generally develop heart
disease later in life compared to men. The moderate
confidence (0.57) points to a noticeable association
between being female and the absence of heart dis-
ease in this group. This could be due to a variety of
factors, including biological differences and healthier
lifestyle choices, such as higher engagement in pre-
ventive healthcare practices among women.
In contrast, in Group 1 (individuals with heart
problems), we observe a similar rule but related to
male individuals: Rule 5 indicates that 57% of in-
dividuals with heart problems are male. This sug-
gests that men in this group are more likely to suf-
fer from heart disease or have had a heart attack.
The confidence value here indicates that being male is
moderately associated with the presence of heart dis-
ease, which is consistent with the previous analysis of
Group 2.
The comparison of these two rules emphasizes the
stark gender differences in heart health between the
groups. In Group 2, being female appears to offer a
protective advantage against heart disease, whereas in
Group 1, being male is more strongly associated with
the presence of heart issues. This highlights the need
for gender-specific approaches in both prevention and
treatment, as women in the general population may
be benefiting from healthier behaviors, while men,
particularly those with heart conditions, may require
more targeted interventions to address lifestyle risks.
5.4 Conflicting Results
Upon analyzing the rules related to cholesterol checks
(CholCheck) across both groups, a potential contra-
diction emerges. Analysing the rules in both groups,
individuals with and without heart problems show a
strong association with cholesterol checks. Specifi-
cally, we observe:
• In Group 1: HasHeartProb → CholCheck and
CholCheck → HasHeartProb.
• In Group 2: NoHeartProb → CholCheck and
CholCheck → NoHeartProb.
At first glance, these rules may seem to present
conflicting outcomes: individuals with heart prob-
lems check their cholesterol, but so do individuals
without heart problems. However, this does not repre-
sent a true contradiction when we consider the differ-
ing contexts in which cholesterol checks occur in each
group. In Group 1, individuals check their cholesterol
as part of a treatment strategy, with cholesterol mon-
itoring being a key component of managing an exist-
ing heart condition. In contrast, in Group 2, choles-
terol checks are undertaken as a preventive measure,
aimed at maintaining heart health and preventing the
onset of heart problems.
In essence, cholesterol monitoring serves dual
purposes: it is both a reactive measure for those man-
aging heart disease and a proactive measure for those
aiming to avoid it.
HEALTHINF 2025 - 18th International Conference on Health Informatics
546

6 CONCLUSION AND FUTURE
WORK
This study utilized Formal Concept Analysis (FCA)
to investigate health indicators associated with heart
disease. From approximately 80 extracted rules, 12
were selected for in-depth analysis, focusing on indi-
viduals with and without heart problems. The anal-
ysis uncovered important patterns, such as the strong
relationship between factors like high cholesterol and
blood pressure with the incidence of heart disease.
In future work, applying this method to larger,
more varied datasets would likely yield richer infor-
mation, particularly when combined with alternative
data processing techniques. Additionally, refining the
binarization process used in this study could help cap-
ture more subtle variations in health behaviors. Ex-
ploring Triadic Concept Analysis (TCA) could also
provide a richer framework by incorporating addi-
tional dimensions such as time or conditions, allow-
ing for a more complex analysis of how health in-
dicators interact over different contexts. This could
further enhance the understanding of the multifaceted
nature of cardiovascular diseases.
Finally, investigating other cardiovascular condi-
tions or expanding the analysis to longitudinal health
data might provide a deeper understanding of how
health indicators evolve over time, potentially im-
proving early detection and prevention strategies.
ACKNOWLEDGEMENTS
The authors thank the Pontif
´
ıcia Universidade
Cat
´
olica de Minas Gerais – PUC-Minas and
Coordenac¸
˜
ao de Aperfeic¸oamento de Pessoal de
N
´
ıvel Superior — CAPES (CAPES – Grant PROAP
88887.842889/2023-00 – PUC/MG, Grant PDPG
88887.708960/2022-00 – PUC/MG - Inform
´
atica,
and Finance Code 001). The present work was also
carried out with the support of Fundac¸
˜
ao de Amparo
`
a Pesquisa do Estado de Minas Gerais (FAPEMIG)
under grant number APQ-01929-22.
REFERENCES
Ananias, K. H., Missaoui, R., Ruas, P. H., Zarate, L. E., and
Song, M. A. (2021). Triadic concept approximation.
Information Sciences, 572:126–146.
Dane, S. and Centers for Disease Control and Prevention
(2015). Behavioral risk factor surveillance system.
Ganter, B. and Wille, R. (2012). Formal Concept Analy-
sis: Mathematical Foundations. Springer Science &
Business Media.
James, S. and Smith, A. (2023). The global burden of car-
diovascular diseases and risk factors in 2023. The
Journal of Cardiology, 89:12–34.
Lancet (2020). Global, regional, and national age-sex-
specific mortality for 282 causes of death in 195 coun-
tries and territories, 1980–2017: a systematic analy-
sis for the global burden of disease study 2017. The
Lancet, 392(10159):1736–1788.
LARIM (2017). Lattice miner. https://github.com/
LarimUQO/lattice-miner.
Miranda, D., Z
´
arate, L., and Song, M. (2024). Application
of formal concept analysis to characterize driving be-
haviors and socio-cultural factors related to driving.
pages 56–62.
Organization, W. H. (2000). Obesity: preventing and man-
aging the global epidemic. Report of a WHO con-
sultation. WHO Technical Report Series 894. World
Health Organization, Geneva.
Organization, W. H. (2021). Cardiovascular diseases (cvds).
Song, M. A. J., Galvez, L. E. Z., and Santos, D. S. (2024).
Applying formal concept analysis to characterize in-
fant mortality: Aplicando an
´
alise de formal de con-
ceitos para a caracterizac¸
˜
ao da mortalidade infantil.
Concilium, 24(3):387–407.
Stumme, G., Taouil, R., Bastide, Y., Lakhal, L., and
Nourine, L. (2002). A survey of formal concept anal-
ysis. In Formal concept analysis: foundations and
applications, pages 123–142. Springer.
Teboul, A. (2022). Heart disease health indicators dataset
notebook.
Wille, R. (1982). Restructuring lattice theory: an approach
based on hierarchies of concepts. In Ordered sets,
pages 445–470. Springer.
ˇ
Skopljanac Ma
ˇ
cina, F. and Bla
ˇ
skovi
´
c, B. (2014). Formal
concept analysis – overview and applications. Proce-
dia Engineering, 69:1258–1267. 24th DAAAM Inter-
national Symposium on Intelligent Manufacturing and
Automation, 2013.
Analysis of Health Indicators for Heart Disease Based on Formal Concept Analysis
547