Application of Artificial Intelligence Technologies and Fuzzy Logic to Practical Problems

Abstract

Artificial intelligence technologies include Fuzzy Logic, Fuzzy Inference System (FIS), Soft Computing, Adaptive Neuro Fuzzy Inference System, Machine Learning, Artificial Neural Networks (ANN), Deep Learning, Condensing Neu-ral Networks (CNN), Generative Adversarial Networks (GAN), Generative Pre-trained Transformer (GPT-3), etc. That is, all these works are included in artificial intelligence. This article mainly analyzed fuzzy logic and fuzzy inference system (FIS). What is the special place of fuzzy logic in artificial intelligence technolo-gies? Fuzzy logic is the mathematics of artificial intelligence technologies. That is, in most of the various fields that we see here, mathematical calculations are solved on a fuzzy basis. Thus, when talking about the mathematics of artificial in-telligence technologies, fuzzy logic and fuzzy set theory come to the fore. What does fuzzy logic do? So what is the advantage of fuzzy logic when there is classi-cal logic? For this, such an issue is considered in this article. According to folk medicine, in order to stay fit, a young person needs to drink 20 g of herbal tea a day, and an old person needs to drink 50 g of herbal tea a day. In this case, how many grams of herbal tea should a person at the age of 45 drink per day? and how many grams of herbal tea should someone not very old drink per day? This is just an example that we mentioned and has nothing to do with reality. It is im-possible to solve this example with classical logic, but fuzzy logic plays an im-portant role in solving this type of problems.

Full text

Azərbaycan Respublikasında 2022-ci ilin “Şuşa İli” elan edilməsi haqqında
Azərbaycan Respublikası Prezidentinin 05 yanvar 2022 tarixli Sərəncamına uyğun olaraq
“Şuşa İli”nəhəsr olunmuş
İNSAN-KOMPÜTER QARŞILIQLI ƏLAQƏSİ
mövzusunda
BEYNƏLXALQ ŞƏRQ KONFRANSININ
MATERİALLARI
INTERNATIONAL EASTERN CONFERENCE ON
HUMAN-COMPUTER INTERACTION
PROCEEDINGS
ISSN 2616-4248
9-10 sentyabr 2022
Naxçıvan, Azərbaycan
IECHCI2022
iechci.info

KONFRANS KOMİTƏLƏRİ
KONFRANSIN SƏDRİ
Dosent Nurlana Əliyeva
Rektor
“Naxçıvan” Universiteti
KONFRANSIN MƏSUL KATİBİ
Şamil Hümbətov
“Naxçıvan” Universiteti
ELM KOMİTƏSİ
AMEA h.ü. Əli Abbasov, İdarəetməSistemləri İnstitutu
Prof. Kürşat Çağıltay, Orta Şərq Texniki Universiteti
Prof. Nergiz Çağıltay, Atılım Universiteti
Prof. Yüksel Göktaş,Atatürk Universiteti
Prof. Engin Kurşun, Atatürk Universiteti
Prof. Hatice Sancar Tokmak, Mersin Universiteti
Prof. Kerem Rizvanoğlu, Galatasaray Universiteti
Dosent Aynur Cəbiyeva, Azərbaycan Dövlət Neft vəSənaye Universiteti
Dosent Azər Hüseyn, Odlar yurdu Universiteti
Dosent KəmaləPaşayeva, Bakı Ali Neft Məktəbi
Dosent Leyla Muradxanlı, Bakı Ali Neft Məktəbi
Dosent Rahib İmamquluyev, Odlar Yurdu Universiteti
Dosent TəranəƏliyeva, Azərbaycan Dövlət İqtisad Universiteti
Dosent Türkan Karakuş Yılmaz, Atatürk Universiteti
Dr. Berkan Çelik, Van Yüzüncü Yıl Universiteti
Dr. Göknur Kaplan, Orta Şərq Texniki Universiteti
Dr. Gulşat Muhametcanova, Qırğızıstan-TürkiyəManas Universiteti
Dr. Nuri Kara, İstanbul Bilgi Universiteti
Dr. Serkan Alkan, Orta Şərq Texniki Universiteti
Dr. Sezin Eşfer Öndünç, Bahçeşehir Universiteti
Dr. Svetlana Əhmədova, Sumqayıt Dövlət Universiteti
Dr. Ayşe Gül Kara Aydemir, Akdeniz Universiteti
Dr. Berrin Doğusoy, Mersin Universiteti

TƏŞKİLAT KOMİTƏSİ
Dosent Nurlana Əliyeva, Sədr
Dr. Hüseyn Bağırsoylu, Sədr müavini
Şamil Hümbətov, Məsul Katib
Dosent AidəMustafayeva, Mingəçevir Dövlət Universiteti
Dr. LətifəAğamalıyeva, Azərbaycan Universiteti
AsudəAbdurrahmanova, Mingəçevir Dövlət Universiteti
Elməddin Hüseynov, Bakı Mühəndislik Universiteti
Türkan Əlibəyli, “Naxçıvan” Universiteti
Rəcəb Cəfərli, “Naxçıvan” Universiteti
İŞÇİ HEYƏT
Şamil Hümbətov, Sədr
Nazlı Əjdərova, “Naxçıvan” Universiteti
Həsən Nəcəfov, “Naxçıvan” Universiteti
Cabir Kərimov, “Naxçıvan” Universiteti
Leyla İbrahimova, “Naxçıvan” Universiteti
Bəşir Bəşirli, “Naxçıvan” Universiteti
Fərhad Heydərov, “Naxçıvan” Universiteti
Oğuz Məmmədov, “Naxçıvan” Universiteti
IECHCI2022 konfransı ali vəorta ixtisas təhsilinin prioritet istiqamətləri üzrəinkişafa xidmət edən
təşəbbüslərin dəstəklənməsi məqsədiləElm vəTəhsil Nazirliyinin “Ali vəorta ixtisas təhsilinin
2021/2022-ci illər üzrəinkişafı” qrant müsabiqəsinin, İxtisaslar üzrəyeni vəmüasir tədris
məzmununun hazırlanması vətətbiqi istiqaməti üzrəqalibi olmuş, “İnsan-Kompüter İnterfeysləri”
sahəsindətəhsil verənlərin peşəkarlığının yüksəldilməsi vəbeynəlxalq təcrübənin öyrənilməsi adlı
qrant layihəsi çərçivəsindəkeçirilmişdir.
TƏRƏFDAŞLAR
HCI.AZ

Security, Trust, Ethics, and AI
Qarşılıqlı Əlaqədə Təhlükəsizlik, Güvən, Etika və Süni İntellekt
202
to light. Some of them are weaker than others due to silicon defects. This can be taken into account if we take an image
of a point source on the optical axis of a camera with a distant lens (which ensures the same illumination of the image
plane), and use the result to correct future brightness measurements.
The article normalized three measurements of the brightness of each image element by dividing each of them by
their sum.
This makes it possible to exclude the effect of unequal sensitivity of the sensor, as well as to take into account
fluctuations in illumination. Moreover, this makes the system insensitive to differences in surface albedo when moving
from one point of the object to another.
By reducing strong noise, imaging device defects, or spots on the surface, the stereophotometric method will not be
able to attribute the orientation of the surface to an isolated image point. We can find such isolated points and assign
them a normal equal to the average of neighboring values. The main reason for doing this is that such a spot can be
mistaken for a hole when calculating the Euler number. To solve the problem of local navigation, a mobile robot can be
equipped with simple ultrasonic or infrared sensors around the perimeter. However, all of the above means cannot give
the robot a complete picture of what is happening around it, since the use of various distance sensors allows the robot to
determine the distance to objects and their dimensions, but does not allow them to determine their other properties -
shape, color, position in space, which leads to to the impossibility of classifying such objects according to any criteria.
ACKNOWLEDGMENT
The author Aynur Jabiyeva is thankful to project manager, Senior Lecturer Shamil Humbatov ("Nakhchivan"
University) for affording scientific assistance to this research work.
REFERENCES
[1] Vorotnikov S. A. Informatsionnye ustroistva robototekhnicheskikh system.[Information devices of robotic systems].Moscow: Izd. MGTU im N.
E. Baumana, 2005. 384 p.
[2] Soifer V. A. (Ed.) Metody komp'yuternoi obrabotki izobrazhenii.[Methods of computer image processing. Moscow: Fizmatlit, 2001. 784 p.
[3] Ostrovskii O. A. Kriminalisticheskaya portretnaya ekspertiza po videoizobrazheniyam, kak forma opredeleniya lichnosti [Forensic portraits on
video images, as a form of identification]. The Eleventh International Conference on Eurasian scientific development Proceedings of the
Conference. Editor Jana Ilyna, Russia. 2016. Pp. 176–179.
[4] Ostrovskii O. A. Printsip ob''ektnoi dekompozitsii v sistematizatsii identifikatsionnykh kodov, kharakterizuyushchikh prestupleniya v sfere
komp'yuternoi informatsii. [The principle of object decomposition in the systematization of indentification on codes characterizing crimes in the
of computer information]. Politseiskaya deyatel'nost' [Police activity].2017. No. 3. Pp.10–18.
[5] Povarkova A. B. Komp'yuternyi analiz izobrazhenii: obshchie svedeniya, sistemy, primery ispol'zovaniya [Computer analysis of images: general
information system, examples of use]. Vestnik infektologii. Elektronnyi zhurnal [Herald of infectious diseases. Electronic journal]. URL:
http://www.infectology. ru/microscopY/today/analvsis/index.aspx (date of the application: 08.09.2017)
[6] Prett U. Tsifrovaya obrabotka izobrazhenii: vdvukhkn [Digital image processing]. In 2nd book. Moscow: Mir, 1982. B. 1. 310 p. , Kn. 2. 790 p.
[7] Rozenfel'd A. Raspoznavanie izobrazhenii [Image Recognition].TIIER. 1981. T. 69. No. 5. Pp. 120–133.
[8] Semenkov O. I, Ablameiko S. V., Bereichik V. I., Starovoitov V. V.Obrabotka i otobrazhenie informatsii v rastrovykh graficheskikh sistemakh
[Processing and display of information in raster graphics systems]. Minsk, Nauka i tekhnika, 1989
Application of Artificial Intelligence
Technologies and Fuzzy Logic to Practical
Problems
Rahib Imamguluyev
Department of IT and Engineering,
Odlar Yurdu University, Baku, AZ1072, Azerbaijan
rahib.aydinoglu@gmail.com
Ilham Hajiyev
Department of IT and Engineering,
Odlar Yurdu University, Baku, AZ1072,
Azerbaijan
ilham.haciyev.97@mail.ru
Abstract — Artificial intelligence technologies include Fuzzy Logic, Fuzzy Inference System (FIS), Soft Computing, Adaptive
Neuro Fuzzy Inference System, Machine Learning, Artificial Neural Networks (ANN), Deep Learning, Condensing Neu-ral
Networks (CNN), Generative Adversarial Networks (GAN), Generative Pre-trained Transformer (GPT-3), etc. That is, all
these works are included in artificial intelligence. This article mainly analyzed fuzzy logic and fuzzy inference system (FIS).
What is the special place of fuzzy logic in artificial intelligence technolo-gies? Fuzzy logic is the mathematics of artificial
intelligence technologies. That is, in most of the various fields that we see here, mathematical calculations are solved on a
fuzzy basis. Thus, when talking about the mathematics of artificial in-telligence technologies, fuzzy logic and fuzzy set theory
come to the fore. What does fuzzy logic do? So what is the advantage of fuzzy logic when there is classi-cal logic? For this,
such an issue is considered in this article. According to folk medicine, in order to stay fit, a young person needs to drink 20 g
of herbal tea a day, and an old person needs to drink 50 g of herbal tea a day. In this case, how many grams of herbal tea
should a person at the age of 45 drink per day? and how many grams of herbal tea should someone not very old drink per
day? This is just an example that we mentioned and has nothing to do with reality. It is im-possible to solve this example with
classical logic, but fuzzy logic plays an im-portant role in solving this type of problems.
Keywords: Artificial intelligence, Fuzzy Logic, Fuzzy Inference System (FIS). Fuzzy set, Soft Computing

Security, Trust, Ethics, and AI
Qarşılıqlı Əlaqədə Təhlükəsizlik, Güvən, Etika və Süni İntellekt
203
I. INTRODUCTION
Artificial Intelligence - The Turing Test An intelligent system is a system that can make decisions in the face of
uncertainty.
The Turing test for determining Artificial Intelligence is a concept first mentioned by the famous English
mathematician and computer scientist Alan Turing in his 1950 article Computing Machinery and Intelligence in the
philosophy journal Mind [1].
Turing test: Suppose we have a wall and behind the wall is a computer and a human. We correspond with them from
the other side of the wall, and the answers to our questions come from the other side [1, 2]. If we cannot choose whether
the answers were given by a person or by a computer, then we say that the computer behind the wall is a system with
enough artificial intelligence (Fig. 1).
Fig. 1. The Turing Test:
Machine Development: Artificial Intelligence
Not by measurement - we perceive by feelings
Not by numerical calculation- Conceptual calculation
1. Example-1: TV sound
Accurate appeal
"Increases TV volume by 45%"
"Reduce the volume of the TV by 15%"
Fuzzy appeal
"Turn up the TV a little"
"Turn down the volume of the TV a little"
"Turn up the volume of the TV a lot"
"Turn up the volume on the TV"
"Turn down the TV volume more"
2. Example-2: Air temprator
By exact calculation
"The temperature of the air is 25 degrees"
"The temperature of the air is -2 degrees"
Fuzzy approach
"The weather is warm"
"It's very hot"
"The weather is cold"
"It's very cold"
Sample from folk medicine
According to folk medicine, a young person should drink 20g of herbal tea per day, and an elderly person should
drink 50g of herbal tea per day.
In such a case, the following questions arise:
a) How many grams of herbal tea should a 45-year-old person drink per day?
b) How many grams of herbal tea should a person not very old drink per day
Solution infrastructure
• Fuzzy logic, Fuzzy set, Fuzzy number [2]
• Fuzzy logic operators
• Linguistic variables
• Operations on linguistic variables
• Fuzzy rules and fuzzy outcome
• Fuzzy inference system (FIS) [3-6]
• Sample calculation

Security, Trust, Ethics, and AI
Qarşılıqlı Əlaqədə Təhlükəsizlik, Güvən, Etika və Süni İntellekt
204
II. APPLICATION OF FUZZY LOGIC
Classical logic:
• Is a person young at 36? NO/YES
• Is it hot at 24 degrees? NO/YES
Fuzzy logic:
• YES to a certain extent [2-8].
Classical logic ()=   Fuzzy logic ()∈[,]
To solve any problem with fuzzy logic, we use numbers in classical mathematics, but here we use linguistic
variables [6-9]. For example, very young, young, middle-aged, old, a little old, etc. The graph of linguistic variables is
shown in the figure below (Fig. 2).
X ==> Age variable
Membership rat e
Linguistic variables
00102030 40 5060708090 100
0.2
0.4
0.6
0.8
1
Middle age
Young Old
Fig. 2. Linguistic variables of age
As you can see in the picture, if X is an age variable, then in classical mathematics x=25, x=35, x=65, etc. happens.
But in fuzzy logic, x=young, x=middle-aged, x=old, etc. happens.
Graphs of elementary and complex linguistic variables are depicted in the figure below (Fig. 3).
00102030 40 50 60 708090100
0.2
0.4
0.6
0.8
1
Young Old
X = Age
00 10 20 30 40 50 60 70 80 90 100
0.2
0.4
0.6
0.8
1
X Age=
Primary Linguistic Values
Composite Linguistic Values
Membership degree Membership degree
Fig. 2. Graphs of elementary and complex linguistic variables.
Operations on linguistic variables:
→( () )= ()
 →( ( ())) =  ()
 → ( ()) =   () (1)
 →  ℎ
2 () , 0 ≤ ()≤0.5
1−2 [1− () ] , 0.5 <  ()≤1
Based on the above formulas, the graphs of the affiliation functions of the linguistic variables young, very young, a
little young, really young are as follows (Fig. 4).

Security, Trust, Ethics, and AI
Qarşılıqlı Əlaqədə Təhlükəsizlik, Güvən, Etika və Süni İntellekt
205
00 10 20 30 40 50 60 70 80 90 100
0.2
0.4
0.6
0.8
1
Very young=CON(Young)
00 10 20 30 40 50 60 70 80 90 100
0.2
0.4
0.6
0.8
1
A little young=D (YoungİL)
00 10 20 30 40 50 60 70 80 90 100
0.2
0.4
0.6
0.8
1
Young=Really young
Fig. 4. Linguistic variables.
There are several standard functions for using linguistic variables. The most commonly used are the following (Fig.
5).
Triangle Trapezoid Sigmoid/exponential Should ered
trapezoid
Very
commo n
shape
Common
shape
Rare
shape Common
shape
Notes: Shapes do not have to b e symmetrical
1 2 34 567891011121314151617
0.5
1.0
Fig. 5. Forms of various functions.
Now, the fuzzy logical inference system (Fuzzy Inference System - FIS) is described in the following figure (Fig. 6).
x is A
1
1
W
1
yB is
1
x is A
2
1
W
2
yB is
2
Qayda 1 (Rule 1)
x is A
r
1
W
r
yB is
r
x
1
(Fuzzy)
(Crisp or
Fuzzy)
(Fuzzy)
(Fuzzy)
Aggregator
(Fuzzy)
Defuzzifi er
(Crisp)
y
Qayda (Rule )22
Qayda (Rule )rr
(kəskin v ə ya
qeyri-səlis)
(Q.səlis)
(Q.səlis)
(Q.səlis) (Q.səlis)
Fig. 6. Fuzzy Inference System - FIS.
Now let's solve a practical problem with fuzzy logic.
Based on years of experience, a rule has been formed among people that young people should drink 20 grams of
herbal tea per day, and elderly people should drink 50 grams of herbal tea per day. In this case, we need to answer the
following two questions
a) How many grams of herbal tea should a 45-year-old person drink per day?
b) How many grams of herbal tea should a person not very old drink per day?
Definition of rules:
Rule-1: If "A young person" then result1=20g
Rule-2: If "An old person" then result2=50g
Definition of terms: ()=1,

 ,
,
<25
∈[25,60]
>60 
()=0,

 ,
,
<40
∈[40,75]
>75 

Security, Trust, Ethics, and AI
Qarşılıqlı Əlaqədə Təhlükəsizlik, Güvən, Etika və Süni İntellekt
206
A= “Not very old” → “Not (Very old)”
()=1−()=1−()=󰇱1,

 ,
,
<40
∈[40,75]
>75 󰇲
The following graphic describes the concepts and membership functions (Fig. 7).
01020 30 40 50 60 70 80 90 100
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
01020 30 40 50 60 70 80 90 100
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
01020 30 40 50 60 70 80 90 100
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
Young
Old
Not oldvery
Fig. 7. Graphical representations of the functions of the belonging of possible term sets on the linguistic variable of age.
Now let's do the calculation based on the FIS result for a 45-year-old person
Rule 1: How young is a person at the age of 45?
(45)=60 −45
35 =0.43
In this case, according to Rule 1, the result should be 0.43 degrees and 20 grams
Rule 2: How old is a person at 45 years old?
ş(45)=45 −40
35 =0.14
In this case, according to Rule 2, the result should be 0.14 degrees and 50 grams.
The aggregated result is the average coefficient of all rules.
FİS(45)=0.43 ∗20 + 0.14 ∗50
0.43 + 0.14 =27.37
So it turned out that a person at the age of 45 needs to drink 27.37 grams of herbal tea per day. We can calculate the
other ages in the same way.
III. REFERENCES
[1] Wayne Patterson, Cynthia E. Winston-Proctor, Turing Tests, In book: Behavioral Cybersecurity, November 2020, DOI:
10.1201/9781003052029-11
[2] L.A. Zadeh, “Fuzzy Sets”, Information and Control 8, 38-53, 1967.
[3] Zadeh, Lotfi A., «Fuzzy Logic, Neural Networks, and Soft Computing», Communications of the ACM, March 1994, Vol. 37 No.
3, pages 77—84.
[4] Lotfi A Zadeh, Rafik Aliev, Fuzzy Logic Theory and Applications: Part I and Part II, https://doi.org/10.1142/10936 | December
2018, Pages: 61.
[5] Zadeh L.A. Toward a theory of fuzzy information granulation and its centrality in human reasoning and fuzzy logic. Fuzz y Sets
and Systems 1997;90:111-127.
[6] Rafik Aliev, Alex Tserkovny, Fuzzy Logic for Incidence Geometry, In book: Beyond Traditional Probabilistic Data Processing
Techniques: Interval, Fuzzy etc. Methods and Their Applications, February 2020, DOI: 10.1007/978-3-030-31041-7_4
[7] Ahmed Valiyev, Rahib Imamguluyev, Gahramanov Ilkin, Application of Fuzzy Logic Model for Daylight Evaluation in
Computer Aided Interior Design Areas, January 2021, In book: 14th International Conference on Theory and Application of
Fuzzy Systems and Soft Computing – ICAFS-2020, DOI: 10.1007/978-3-030-64058-3_89
[8] Tarlan Abdullayev, Rahib Imamguluyev, Niyar Umarova, Application of Fuzzy Logic Model for Optimal Solution of Light
Reflection Value in Lighting Calculations, January 2022, In book: 11th International Conference on Theory and Application of

Security, Trust, Ethics, and AI
Qarşılıqlı ƏlaqədəTəhlükəsizlik, Güvən, Etika vəSüni İntellekt
207
Soft Computing, Computing with Words and Perceptions and Artificial Intelligence - ICSCCW-2021 DOI: 10.1007/978-3-030-
92127- 9_53
[9] Rahib Imamguluyev, Application of Fuzzy Logic Model for Correct Lighting in Computer Aided Interior Design Areas, January
2021, In book: Intelligent and Fuzzy Techniques: Smart and Innovative Solutions, DOI: 10.1007/978-3-030-51156-2_192
Financial Fraud Detection using various
Machine and Deep Learning techniques for
better performance
Aytaj Abdullayeva
Process Automation Engineering Department
Baku Higher Oil School
Baku, Azerbaijan
Aytaj.abdulayeva.std@bhos.edu.az
Kamala Pashayeva
Process Automation Engineering Dep
artment
Baku Higther Oil School
Baku, Azerbaijan
Kamala.pashayeva@bhos.edu.az
Abstract —The increase of fraudulent actions in the last decade caused immense amount of money loss for not only lots of
companies, but also agencies and organizations. The modern techniques can detect frauds and take an action, while
traditional methods can be time, resource consuming and not effective, as humans cannot catch the hidden patters on the
data for the correct predictions. The suggested method on this paper is related to using ensemble method to merge three
higher score model for getting better performance and using user identifications for defining not only fraudulent actions, but
also the frauds.
Keywords —Machine Learning, Deep Learning, Fraud Detection, Automatic Fraud Detection
I. INTRODUCTION
The rate of digital payments is increasing tremendously all over the world. Hence, the number of transactions
handled by the financial companies is on the rise, for example, in the fourth quarter of the 2021, PayPal processed
approximately 5.5 billion transactions that is 22% greater than that of 2020. The number of the loss caused by frauds
has increased immensely having the number of $33 billion in 2021, with 6.94 cents for every $100. According to the
study of “VynZ Research”, it is stated that the market value of the fraud detection and prevention is expected to reach
$86 billion by the end of 2025 [[1]].
Automatic detection of the financial frauds is an integral part of the work done by banks or financial institutions. Of
course, it is very vital to integrate this automatic approach for detecting fraudulent transactions correctly with highest
possible performance. In order to tackle this problem, Machine Learning, which is the main part of AI (Artificial
Intelligence), will be applied, it learns from the experience and makes predictions about new datasets of the financial
institutions. Recent studies show that false declines of the payments caused $118 billion in 2021 for the merchants [[2]].
Building model for correctly determining fraud is not an easy task to do, it requires correct selection of strategy,
algorithms, the selection of the features that should be included to the model and, of course, solution to deal with the
class imbalance of target because most of the ML models underperform under this circumstance. Of course, there are
several limitations due to the provided dataset because all user information, as it is publicly available dataset, is masked
due to the security reasons. The last and most crucial step is correctly defining the evaluation metrics and compare the
models based on that (See Fig. 1).
Figure I. Process flowchart for financial fraud detection