Fuzzy Logic for Diabetes Predictions: A Literature Review

Alice Tissot Garcia Pintanel

1,∗

, Grac¸aliz Pereira Dimuro

, Eduardo Nunes Borges

, Giancarlo Lucca

and Camila Rose Guadalupe Barcelos

Computational Modeling, Federal University of Rio Grande (FURG), Rio Grande, RS, Brazil

Center for Computational Sciences (C3), Federal University of Rio Grande (FURG), Rio Grande, RS, Brazil

Hospital Sirio Libanes, S

ao Paulo, SP, Brazil

Keywords:

Diabetes, Machine Learning, Classiﬁcation Problems, Systematic Literature Review.

Abstract:

The use of methodologies based on machine learning is being increasingly used in health systems today,

addressing different areas such as food, society, health and others. In terms of health, different techniques were

applied to classify different diseases. In this sense, diabetes is an important and silent disease that deserves

special attention and care. Individuals often do not know they have it, and, therefore, seeking alternatives to

predict this disease is an important contribution to the health area. Thinking about it, in this work we present

a systematic review of the literature with the objective of observing which strategies are currently being used

to predict and classify diseases using fuzzy logic, in particular, diabetes. For this, 6 works were selected and

analyzed, where the technique for obtaining the considered information is the blood test, in order to understand

the current state of the art.

1 INTRODUCTION

Health has always been and will be an extremely im-

portant subject to be debated. Either through the in-

cessant search for new methods that can mitigate the

risks of getting diseases, or through the search for

new strategies that can somehow contribute to the im-

provement of the health system.

The use of machine learning techniques (Bonac-

corso, 2017) has been increasingly used in several

areas of society, including medicine. For example,

in (Saleck et al., 2017) the authors used a cluster-

ing algorithm to detect tumors in mammography im-

ages. In (Bergquist et al., 2017), the authors used ma-

chine learning to classify lung cancer patients receiv-

ing chemotherapy into early-stage versus advanced-

stage cancer.

Diabetes is a disease caused by insufﬁcient pro-

duction or poor absorption of insulin, a hormone that

regulates blood glucose and provides energy for the

body (Grillo and Gorini, 2007). Diabetes can cause

increased blood glucose and can lead to complica-

tions in the heart, arteries, eyes, kidneys and nerves.

In more severe cases, diabetes can even lead to death.

According to the Brazilian Society of Diabetes

∗

PhD student

https://diabetes.org.br/

at the end of 2021, Brazil had more than 13 mil-

lion people living with diabetes, which represents

about 6.9% of the Brazilian population. Diabetes can

present itself in different ways and has several dif-

ferent types, with type 2 diabetes being the one that

most affects people, about 90% of people with dia-

betes have type 2. Depending on the severity, dia-

betes can be controlled with physical activity and di-

etary planning, but in other cases, it requires the use

of insulin and/or other medications to control glucose

(Lyra et al., 2006).

One of the tasks most used by machine learning

techniques refers to the prediction problem. Predic-

tive analysis basically consists of applying algorithms

to understand the existing data structure and gener-

ate ways to help predict new cases. According to

(Dos Santos et al., 2019), in the health area, predictive

models can be used to estimate the risk of a given out-

come occurring, considering a set of characteristics,

such as: socioeconomic and demographic aspects, re-

lated to lifestyle and health conditions, among oth-

ers. In addition, the correct use of prediction models

can result in positive implications in terms of cost re-

duction and the effectiveness of interventions, such as

treatments and preventive actions.

Considering the severity and consequences that

diabetes can have on people’s lives, it was understood

476

Pintanel, A., Dimuro, G., Borges, E., Lucca, G. and Barcelos, C.

Fuzzy Logic for Diabetes Predictions: A Literature Review.

DOI: 10.5220/0011851500003467

In Proceedings of the 25th International Conference on Enterprise Information Systems (ICEIS 2023) - Volume 1, pages 476-483

ISBN: 978-989-758-648-4; ISSN: 2184-4992

 2023 by SCITEPRESS – Science and Technology Publications, Lda. Under CC license (CC BY-NC-ND 4.0)

that monitoring the variables involved in the detection

of diabetes could be a way of mitigating the risks for

those people who already have the disease, or even

prevent the disease from manifesting itself in new

people. In addition, the need to control several vari-

ables that can aggravate or even cause diabetes also

led to the need not only to monitor, but also to control

such variables.

For this, however, it was necessary to know the

state of the art with regard to methodologies used in

the health area in the classiﬁcation and prediction pro-

cess that could help in understanding the problem and

in the elaboration of a new methodology.

In order to try to map which techniques are being

most used in the prediction of diseases and, mainly,

in the prediction of diabetes, it was decided to carry

out a systematic review, so that one could ﬁnd what is

currently state-of-the-art in what concerns related to

disease prediction.

In addition, the idea is also to ﬁnd, from the sys-

tematic review, how fuzzy logic (Zadeh et al., 1996)

is being used in this area, since it is an approach that

uses linguistic terms and therefore manages to re-

turn interpretable results. We pointed out that fuzzy

logic allows the possibility of modeling such linguis-

tic terms according to the domain expert, which can

be an advantage over other methods. Therefore, fuzzy

logic based questions were included in this review.

Then, a systematic review of the literature was car-

ried out, in 4 bibliographical sources (BS), consider-

ing 4 research questions.

The general objective of this work is to iden-

tify the prediction and classiﬁcation methodologies,

mainly involving fuzzy logic, which are being used in

the health ﬁeld, in particular, in relation to diabetes.

The rest of the article is divided into four sec-

tions. The ﬁrst presents the deﬁnition and steps to

build a systematic review; the second presents the

methodology adopted for carrying out the system-

atic research, as well as the discussion in each of the

stages; the third presents the obtained results; and the

last presents the conclusions obtained from the con-

struction of this work.

2 SYSTEMATIC REVIEW

According to (Cordeiro et al., 2007), the System-

atic Literature Review (SLR) consists of a research

method that aims to identify, evaluate and interpret all

relevant research on speciﬁc issues or area of interest.

From the use of the SLR it is possible to sort and syn-

thesize studies from the deﬁnition of questions and/or

keywords, making the review much fairer.

A research protocol must be deﬁned and must con-

tain four generic steps: the ﬁrst one is responsible

for deﬁning the research questions that the system-

atic review intends to answer and that will serve as a

guideline for the study; the second deﬁnes the search

strategy, including the databases and search terms that

will be considered to identify and select the articles;

the third deﬁnes the inclusion and exclusion criteria;

and the fourth and last, is responsible for deﬁning how

to characterize the studies, which information will be

extracted and how it will be synthesized and analyzed

(Keele et al., 2007) (Ercole et al., 2014).

The main advantage with regard to carrying out a

systematic review, according to (Ercole et al., 2014),

consists in the fact that a well-deﬁned methodology

ends up making it less likely that the results of the

literature are biased, that is, it tends to prevent articles

from chosen by the writer are only those who support

his/her point of view.

3 METHODOLOGY

The methodology used to carry out the systematic re-

view followed the method presented by (Keele et al.,

2007), and is summarized in the ﬂowchart presented

by Figure 1. In it, as the process evolves, the articles

found in searches are removed from the scope of the

study. Below, for each one of the four keys stages

of the systematic review process, the adopted criteria

will be discussed in detail.

3.1 Research Questions

The ﬁrst step, to start the systematic review process,

is related with the deﬁnition the research questions

necessarily to be answerd. Such questions are re-

sponsible for giving rise to the search terms and com-

binations of keywords used during searches in the

databases used.

Considering the problem previously discussed in

the Introduction, four research questions were sepa-

rated, namely:

• Question 1: What are the predictive methods

used to adjust models on different blood labora-

tory test variables?

• Question 2: Which classiﬁcation methods are

used to ﬁt models on different blood laboratory

test variables?

• Question 3: What are the predictive meth-

ods used to adjust models on different health

databases?

Fuzzy Logic for Diabetes Predictions: A Literature Review

477

Research Questions

Search Terms

Data

base

Books or

Chapters

Articles

Published

until 2017

Published

after 2017

Year Title

Yes

Relevant

Abstract

Yes

Relevant

Introduction

Conclusion

NoNo

Yes

Relevant

Figure 1: Flowchart of the theoretical review methodology (Source: Elaborated by the author).

• Question 4: Which classiﬁcation methods are

used to ﬁt models on different health databases?

It should be noted that for each of the planned

surveys, the existing fuzzy logic related to predictive

methods was also analyzed, since the idea of the work

is to use fuzzy logic to predict diabetes based on in-

formation extracted from blood tests.

3.2 Search Terms

The search terms were deﬁned from the four research

questions previously presented. We highlight the fact

that we chose to consider the search terms only in En-

glish.

The search terms considered for the construction

of combinations of keywords were: blood test, health

database, classiﬁcation, prediction and fuzzy. From

these selected search terms, some combinations of

keywords were elaborated that were used in the sys-

tematic research in the selected databases. In Table 1,

the relation of keywords (K) used are presented.

3.3 Considered Data Bases

To perform a complete and more robust search, dif-

ferent digital libraries were considered. Precisely, the

queries were performed in:

• IEEE Xplore

;

https://ieeexplore-ieee-

Table 1: Keyword combinations used.

K Combination

1 “blood test” and “prediction”

2 “blood test” and “classiﬁcation”

3 “blood test” and “prediction” and “fuzzy”

4 “blood test” and “classiﬁcation” and “fuzzy”

5 “health database” and “prediction”

6 “health database” and “classiﬁcation”

7 “health database” and “prediction” and “fuzzy”

8 “health database” and “classiﬁcation” and “fuzzy”

• ACM Digital Library

;

• Scopus

;

• SpringerLink

3.4 Exclusion/Inclusion Criteria

Exclusion Criteria (EC) are deﬁned as aspects of the

study that meet the inclusion criteria, but have addi-

tional characteristics that could interfere with the suc-

cess of the study or increase the risk of obtaining un-

necessary information for the study. The considered

ECs were:

• EC 1: Articles published before 2017, because

org.ez40.periodicos.capes.gov.br/Xplore/guesthome.jsp

https://dl-acm-org.ez40.periodicos.capes.gov.br/

https://www-scopus.ez40.periodicos.capes.gov.br/

search/form.uri?display=basic#basic

https://link-springer-com.ez40.periodicos.capes.gov.br/

ICEIS 2023 - 25th International Conference on Enterprise Information Systems

478

Table 2: Total articles returned for each of the steps described.

BS Initial EC 1 EC 2 e 4 EC 3

K1: 30 K5: 4 K1: 20 K5: 4 K1: 18 K5: 4 K1: 18 K5: 4

K2: 32 K6: 1 K2: 20 K6: 1 K2: 19 K6: 1 K2: 19 K6: 1

IEEE Xplore K3: 2 K7: 0 K3: 2 K7: 0 K3: 2 K7: 0 K3: 2 K7: 0

K4: 2 K8: 0 K4: 1 K8: 0 K4: 1 K8: 0 K4: 1 K8: 0

Total: 71 Total: 48 Total: 45 Total: 45

K1: 9251 K5: 596 K1: 5052 K5: 374 K1: 2902 K5: 190 K1: 2892 K5: 190

K2: 17494 K6: 1355 K2: 8812 K6: 766 K2: 5549 K6: 527 K2: 5345 K6: 526

SpringerLink K3: 861 K7: 106 K3: 624 K7: 81 K3: 123 K7: 9 K3: 123 K7: 9

K4: 991 K8: 116 K4: 679 K8: 87 K4: 143 K8: 13 K4: 143 K8: 13

Total: 30770 Total: 16475 Total: 9456 Total: 9241

K1: 1321 K5: 67 K1: 698 K5: 44 K1: 672 K5: 41 K1: 636 K5: 41

K2: 974 K6: 148 K2: 488 K6: 75 K2: 475 K6: 65 K2: 441 K6: 65

Scopus K3: 7 K7: 0 K3: 6 K7: 0 K3: 6 K7: 0 K3: 6 K7: 0

K4: 7 K8: 1 K4: 2 K8: 0 K4: 2 K8: 0 K4: 2 K8: 0

Total: 2525 Total: 1313 Total: 1261 Total: 1191

K1: 92 K5: 13 K1: 66 K5: 6 K1: 45 K5: 4 K1: 45 K5: 4

K2: 152 K6: 18 K2: 69 K6: 6 K2: 49 K6: 5 K2: 49 K: 5

ACM K3: 16 K7: 2 K3: 12 K7: 0 K3: 7 K7: 0 K3: 7 K7: 0

K4: 23 K8: 3 K4: 12 K8: 0 K4: 8 K8: 0 K4: 8 K8: 0

Total: 319 Total: 171 Total: 118 Total: 118

K1: 10694 K5: 680 K1: 5836 K5: 428 K1: 3637 K5: 239 K1: 3591 K5: 239

Total by K K2: 18652 K6: 1522 K2: 9389 K6: 848 K2: 6092 K6: 598 K2: 5854 K6: 597

K3: 886 K7: 108 K3: 644 K7: 81 K3: 138 K7: 9 K3: 138 K7: 9

K4: 1023 K8: 120 K4: 694 K8: 87 K4: 154 K8: 13 K4: 154 K8: 13

the idea is to observe what is newest being de-

veloped and studied by scholars and researchers

in the area;

• EC 2: Books and book chapters, because the idea

is to ﬁnd what is being developed by other re-

searchers in recent years and not the concepts and

deﬁnitions related to the subject addressed in the

work;

• EC 3

: Works that are not in English;

• EC 4: Works that are not fully published. Ex.

drafts;

• EC 5: Articles returned in combinations that do

not involve fuzzy logic.

Inclusion Criteria (IC) are deﬁned as the key char-

acteristics of the population or research being carried

out. The ICs are used to answer the research questions

and are presented below:

• Title: Does the paper deal with machine learning

with diabetes or prediction or classiﬁcation using

fuzzy logic for diabetes?

• Abstract: Does the article explain the use of

fuzzy logic in predicting blood test data? Does

the paper explain about the use of fuzzy logic in

classifying blood test data? Does the paper de-

scribe methods of predicting or classifying blood

test data? Does the paper present prediction or

classiﬁcation in a blood test database?

• Introduction and Conclusion: Are the objec-

tives clearly deﬁned? Are the prediction or classi-

ﬁcation methods clearly deﬁned? Does the study

present the idea of using fuzzy logic in predicting

or classifying blood tests? Does the study present

the blood test database used?

3.5 Selecting Works

In this subsection, the data related to exclusion and

inclusion criteria are presented, as well as the number

of articles that were returned from the application of

each of the previously listed criteria.

The ﬁrst step consisted of carrying out a search

for each of the previously highlighted keyword in the

considered databases, without considering, at ﬁrst,

any exclusion criteria. In Table 2, the total number

of articles returned for each search order. By line, the

digital libraries are considered and by columns, the

stages of the review process (see Figure 1).

Then, it was decided to apply the exclusion crite-

rion related to the year of publication of the returned

results. For this case, therefore, only the results pub-

lished from 2017 to the present were considered. This

exclusion criterion reduced the total number of re-

turned results, as shown in column “EC 1” of Table

The next steps consisted of applying the other ex-

clusion criteria. The column “EC 2 e 4” of Table 2

presents the results after excluding books and book

Fuzzy Logic for Diabetes Predictions: A Literature Review

479

chapters and also articles that were not fully pub-

lished, while the column “EC 3” of Table 2 presents

the results after the application of the EC referring to

the language in which the article was published.

In short, Table 2 presents for each of the biblio-

graphic sources the total number of articles returned

for each K, as well as the total number of articles re-

turned per database. In addition, through Table 2 it

can be observed that with each EC applied, the total

number of articles returned, not only by K, but also

by bibliographic source, decreased.

After evaluating the returned results, it was found

that a very large number of results had been returned

in combinations of words that did not include the

word ”fuzzy”. Thus, it was decided to remove such

combinations from the analysis, since the main focus

of this work was to ﬁnd studies that used fuzzy logic

in the process of predicting or classifying health prob-

lems. Thus, instead of 8 Ks, only 4 were considered,

as shown in Table 3.

Table 3: Total articles selected from the application of all

exclusion criteria.

IEEE

Xplore

Springer

Link

Scopus

ACM

Dig.

Lib.

Total

3 2 123 6 7 138

4 1 143 2 8 154

7 0 9 0 0 9

8 0 13 0 0 13

Total 3 288 8 15 314

Analyzing Table 3, it can be seen that the IEEE

Xplore library was the one that returned a smaller

number of articles considering all Ks. On the other

hand, SpringerLink was the library that returned the

highest number of results. In addition, the K that re-

turned the most articles was K4, which considered the

relationship of “blood test” and “classiﬁcation” and

“fuzzy”.

After applying all the previously listed exclusion

criteria, the next step is to apply the inclusion crite-

ria. In this case, the title of all pre-selected works is

initially evaluated. If the title did not meet the es-

tablished criteria, the article was rejected, if it did, it

moved on to the next analysis. After checking the title

of the 314 articles that had remained, 124 articles still

remained. It is noteworthy that, at that time, they were

not yet being analyzed whether the same article was

appearing in more than one combination or in more

than one database.

The next step consisted of analyzing the abstract

and the idea was similar to that of the title, if the

abstract did not present the desired information, the

work was discarded, otherwise, the introduction and

conclusion of the work were evaluated. Figure 2

presents the summary of the number of articles ex-

cluded from the application of each exclusion and in-

clusion criteria considered in this systematic review.

Therefore, after applying all the inclusion criteria,

the total number of selected articles was equal to 6,

whose title information, authors and year of publica-

tion are presented in Table 4.

Next, the 6 selected articles will be brieﬂy dis-

cussed, using each of the considered research ques-

tions as a guide. It should also be noted that a certain

study can answer one or more questions.

4 RESULTS

In this section, each of the selected articles is brieﬂy

presented, highlighting the questions that each one

of them answers, as well as highlighting their main

points.

However, before starting such an analysis, it was

decided to create a word cloud from the six selected

articles. The cloud was generated from the full text

of all articles, that is, all sections of each article were

considered and not just any speciﬁc one. The purpose

of creating a word cloud is to observe which words

appear more frequently in the analyzed articles. For

the construction of the cloud presented by Figure 3 the

words/terms that appear more than ﬁfty times were

considered.

Thus, it can be concluded that the words that ap-

pear most frequently are precisely those words that

form the scope of study of this work, that is, those

words that formed the queries. In addition to these,

words such as patients and diabetes also appear fre-

quently, since even though they were not part of the

queries, they were terms that were part of the general

scope of the work and were even used as inclusion

criteria for the selection of articles.

The study presented by (Goldar et al., 2020) arises

as an answer to research questions 1 and 2, since it

works with a prediction system for laboratory tests.

The prediction techniques used were Takagi-Sugeno

zero order fuzzy modeling and sequential direct se-

lection method. The authors proposed a prediction

approach in order to predict values of laboratory tests

of patients admitted to the ICU with gastrointestinal

hemorrhage.

A study that helps answer research question 2

and 4 is presented (Wedagu et al., 2020). The

study disregards the question of blood tests, since

it aims to propose a recommendation method called

DIMERS (Diabetes Medicine Recommendation Sys-

tem, which combines a previous medical knowl-

ICEIS 2023 - 25th International Conference on Enterprise Information Systems

480

71 Citations identified

by searching the

IEEE Xplore database

30770 Citations

identified by searching

the Springer Link

database

2525 Citations

identified by

searching the Scopus

database

319 Citations

identified by

searching the ACM

database

33685 Records

screened

10810 Full-text

articles assessed

for eligibility

22875 Records excluded

- 15678 CE 1

- 7127 CE 2 e 4

- 70 CE 3

10496 Records

excluded referring to

queries 1, 2, 5 and 6

314 Full-text

articles assessed

for eligibility

308 Full-text articles excluded

- 124 title

- 102 abstract

- 82 introduction and conclusion

6 Studies included

in the systematic

review

Figure 2: PRISMA Flow Diagram (Source: Elaborated by the author).

edge of doctors with bidirectional long-term memory

(BiLSTM), in order to correctly recommend medi-

cations for diabetes. Also, according to this study,

more than 250,000 people die from medication errors,

which makes it necessary to use machine learning ap-

proaches to correct such problems. However, most of

the approaches adopted do not consider prior medical

knowledge, meaning that they do not have a result as

satisfactory as what is being proposed in the study.

The study presented by (Deif et al., 2021), despite

not meeting the inclusion criteria IC1, was considered

as it aims to use an Adaptive Neuro-Fuzzy Inference

System (ANFIS) to quickly detect cases of COVID-

19 from commonly available laboratory tests . Thus,

this study answers research question 1, since it uses

the fuzzy methodology to predict, from blood tests,

whether or not the patient has COVID-19.

Fasting blood glucose is considered one of the

most important indicators of diabetes, but its testing

is not feasible for the public and requires prior prepa-

rations before implementation. Thus, the study pre-

sented by (Faraji-Biregani and Nematbakhsh, 2019)

aims to present a model for predicting fasting blood

sugar from other factors in blood tests. For this,

the sine-cosine optimization algorithm and neural net-

works (RNN) are used to perform the prediction. The

study was the one of that best answered the research

questions, being able to answer the four questions si-

multaneously.

Different machine learning techniques, such as the

Adaptive Neuro-Fuzzy Inference System (ANFIS),

K-Nearest Neighbors (KNN) and Decision Tree (DT)

were applied in (Kalaiselvi et al., 2022). Such tech-

niques can help in decision making of the physician’s

decision in the process of predicting liver disease.

This study answers questions 1 and 2, since, based

on blood tests, it uses machine learning techniques to

help physicians predict and classify liver diseases. In

addition, one of the techniques presented by the study

makes use of fuzzy logic, which is also of interest to

be analyzed. Together with the study presented in the

previous paragraph, this work was classiﬁed as one

that best answered the listed research questions.

In (Jamuna and Mohan Kumar, 2020), it is pre-

Fuzzy Logic for Diabetes Predictions: A Literature Review

481

Table 4: Basic information of selected articles.

Title Authors Year Total Ci-

tations

Diabetes Fuzzy

Predicting lab values for gastroin-

testinal bleeding patients in the in-

tensive care unit: A comparative

study on the impact of comorbidi-

ties and medications

Mahani, Golnar K and Pa-

joohan, Mohammad-Reza

2019 2 No Yes

Medicine Recommendation System

For Diabetes Using Prior Medical

Knowledge

Wedagu, Mulubrhan Ay-

alew and Chen, Dehua and

Hussain, Muhammad Ather

Iqbal and Gebremeskel,

Tsegay and Orlando,

Mayugi Tanguy and Man-

zoor, Arslan

2020 0 Yes No

Adaptive neuro-fuzzy inference

system (ANFIS) for rapid diagnosis

of COVID-19 cases based on

routine blood tests

Deif, Mohanad and Ham-

mam, Rania and Solyman,

Ahmed

2021 9 No Yes

Diabetes Prediction Recommender

System based on Artiﬁcial Neu-

ral Networks and Sine-Cosine Op-

timization Algorithm

Faraji-Biregani, Maryam

and Nematbakhsh, Nasser

2019 1 Yes No

Liver Disease Prediction Using Ma-

chine Learning Algorithms

Kalaiselvi, R and Meena, K

and Vanitha, V

2021 1 No Yes

Prediction of Diabetes and Cluster-

ing Based on its Levels using Fuzzy

C Means Algorithm

S. Jamuna, K. Mohan Ku-

mar

2020 1 Yes Yes

Figure 3: Word cloud assembled from the six selected arti-

cles (Source: Elaborated by the author).

sented a study that aims to classify patients accord-

ing to three levels of diabetes: non-diabetic, pre-

diabetic or diabetic. For this, it makes use of the

Fuzzy C−Means algorithm. The information used by

the algorithm comes from blood tests of patients. Like

the last two works, this study was also considered one

of the three that best answered the research questions,

since it answers all the items of research question 2,

since it is based on blood test data, uses fuzzy tech-

niques, performing the classiﬁcation of patients ac-

cording to the level of diabetes.

Finally, we highlight the fact that the questions

that were best answered are related to the prediction

and classiﬁcation methodologies used in the health

area. In summary, it can be concluded that of the 6 se-

lected articles: 4 of them answered research question

1, 5 papers answered question 2, 1 answered question

3 and 2 answered question 4.

5 CONCLUSION

Methodologies based on machine learning are in-

creasingly being used to solve problems associated

with healthcare. Diabetes is a disease that needs spe-

cial care because it is a silent disease, making many

ICEIS 2023 - 25th International Conference on Enterprise Information Systems

482

individuals not even know they have such a disease.

Considering the problems mentioned above, the

objective of this work is to identify forecasting and

classiﬁcation methodologies, mainly involving fuzzy

logic, used to help the health area.

To this end, a systematic review was carried out

with the objective of ﬁnding prediction and classiﬁca-

tion methods for diabetes or even for other diseases,

but that could help in understanding the state of the

art, that is, that could facilitate the understanding of

which techniques, currently, are being used to predict

or classify health problems.

In order to carry out the systematic review, inclu-

sion and exclusion criteria were considered so that the

works found could be ﬁltered and, at the end, those

that best met the pre-established conditions could be

studied.

Thus, at the end of the work, it was possible to

observe that techniques using fuzzy logic, such as:

Fuzzy C−Means, ANFIS and Takagi-Sugeno zero or-

der fuzzy modeling are being used and from that it

can be observed which is the best way to contribute

in the area proposing techniques that can improve

the quality/perfomance/explainability/interpretability

of the existing related methods.

The next step, therefore, consists of implementing

a fuzzy methodology for predicting diabetes based on

blood test variables and also on information provided

by the patients themselves, such as: sleep, physical

activity and diet.

ACKNOWLEDGEMENTS

This work is supported by the Academic Master’s

and Doctorate Program for Innovation of the National

Council for Scientiﬁc and Technological Develop-

ment CNPq.

Public Notice FAPERGS/CNPq 07/2022 - Pro-

gram to Support the Settlement of Young Doctors in

Brazil

REFERENCES

Bergquist, S. L., Brooks, G. A., Keating, N. L., Landrum,

M. B., and Rose, S. (2017). Classifying lung can-

cer severity with ensemble machine learning in health

care claims data. In Machine Learning for Healthcare

Conference, pages 25–38. PMLR.

Bonaccorso, G. (2017). Machine learning algorithms.

Packt Publishing Ltd.

Cordeiro, A. M., Oliveira, G. M. d., Renter

ıa, J. M., and

Guimar

aes, C. A. (2007). Revis

ao sistem

atica: uma

revis

ao narrativa. Revista do Col

egio Brasileiro de

Cirurgi

oes, 34:428–431.

Deif, M., Hammam, R., and Solyman, A. (2021). Adaptive

neuro-fuzzy inference system (anﬁs) for rapid diagno-

sis of covid-19 cases based on routine blood tests. Int.

J. Intell. Eng. Syst, 14(2):178–189.

Dos Santos, H. G., do Nascimento, C. F., Izbicki, R.,

de Oliveira Duarte, Y. A., and Chiavegatto Filho, A.

D. P. (2019). Machine learning for predictive analy-

ses in health: an example of an application to predict

death in the elderly in s

ao paulo, brazil. Cadernos de

saude publica, 35(7):e00050818.

Ercole, F. F., Melo, L. S. d., and Alcoforado, C. L. G. C.

(2014). Revis

ao integrativa versus revis

ao sistem

atica.

Revista Mineira de Enfermagem, 18(1):9–12.

Faraji-Biregani, M. and Nematbakhsh, N. (2019). Dia-

betes prediction recommender system based on artiﬁ-

cial neural networks and sine-cosine optimization al-

gorithm. In 2019 5th Conference on Knowledge Based

Engineering and Innovation (KBEI), pages 263–268.

IEEE.

Goldar, S. Z., Ghiasi, A. R., Badamchizadeh, M. A., et al.

(2020). An anﬁs-pso algorithm for predicting four

grades of non-alcoholic fatty liver disease. In 2020

International Congress on Human-Computer Interac-

tion, Optimization and Robotic Applications (HORA),

pages 1–5. IEEE.

Grillo, M. d. F. F. and Gorini, M. I. P. C. (2007).

Caracterizac¸

ao de pessoas com diabetes mellitus tipo

2. Revista Brasileira de Enfermagem, 60:49–54.

Jamuna, S. and Mohan Kumar, K. (2020). Prediction of

diabetes and clustering based on its levels using fuzzy

c means algorithm. International Journal of Scientiﬁc

and Technology Research, 9(2):3222–3225.

Keele, S. et al. (2007). Guidelines for performing system-

atic literature reviews in software engineering. Tech-

nical report, Technical report, ver. 2.3 ebse technical

report. ebse.

Lyra, R., Oliveira, M., Lins, D., and Cavalcanti, N.

(2006). Prevenc¸

ao do diabetes mellitus tipo 2. Ar-

quivos Brasileiros de Endocrinologia & Metabologia,

50:239–249.

Saleck, M. M., ElMoutaouakkil, A., and Mouc¸ouf, M.

(2017). Tumor detection in mammography images us-

ing fuzzy c-means and glcm texture features. In 2017

14th International Conference on Computer Graph-

ics, Imaging and Visualization, pages 122–125.

Wedagu, M. A., Chen, D., Hussain, M. A. I., Gebremeskel,

T., Orlando, M. T., and Manzoor, A. (2020). Medicine

recommendation system for diabetes using prior med-

ical knowledge. In Proceedings of the 2020 4th In-

ternational Conference on Vision, Image and Signal

Processing, pages 1–5.

Zadeh, L. A., Klir, G. J., and Yuan, B. (1996). Fuzzy sets,

fuzzy logic, and fuzzy systems: selected papers, vol-

ume 6. World Scientiﬁc.

Fuzzy Logic for Diabetes Predictions: A Literature Review

483