Applications of AI

1
Applications of Artificial Intelligence to Real World Problems
In recent years, the advancement in logic and theoretical computer science has
encouraged the researchers to understand the importance of Artificial Intelligence (AI). Artificial
Intelligence has been applied to many areas such as military, industry, science, and
manufacturing.
The term ‘Artificial Intelligence’ has become very popular in recent years, but the actual
work on Artificial Intelligence started in the early 1950s. Several researchers have different
opinions on the origin of Artificial Intelligence. According to Munakata (2008), “Artificial
Intelligence is evolved to replace human intelligence with the machine during the industrial
revolution started around 1760” (p.5). This research was emphasized in the early development of
Artificial Intelligence. This finding is in contrast with other research that has been done. Black
and Ertel (2011), states that “Foundation for logic and theoretical computer science laid by
Godel, Church, and Turning in 1930s created interest in Artificial Intelligence” (p.5). He also
states that “the first experimental research conducted by Turning is lead to the invention of
Artificial Intelligence” (p.6). Another researcher Jackson (1985), also supports the experiment
conducted by Turing. He states that “The classic experiment proposed for determining whether a
machine possesses intelligence on a human level is known as Turing’s test” (p.2). These two
findings are crucial because it demonstrated the experiment conducted to identify the human
behavior which in turn lead to invention of Artificial Intelligence. In another study, Swarup
(2012) argues that the “evidence of Artificial Intelligence can be tracked back to ancient Egypt,
but with the development of electronic computer in 1941” (p.2). He also states that “the term
‘Artificial Intelligence’ was first coined by John McCarthy in 1956, at the Dartmouth

2
conference” (p.3). This research gives the evidence of when the term Artificial Intelligence was
used first. These findings provided the origin and the history of Artificial Intelligence.
What is “Artificial Intelligence”? According to Jackson (1985) “Artificial Intelligence is
the ability of the machine to do things that people would say require intelligence” (p.1). This
study provides the basic understanding and definition of the Artificial Intelligence. A few more
authors Russell and Norvig (2003) support this claim, saying “AI definitions vary along two
main dimensions. This study provides two dimensions of Artificial Intelligence. The first one is
mainly concerned with thought processes and reasoning, whereas the second one on the bottom
address behavior” (p.1). This finding helped to measure the human performance and ideal
concept of intelligence. However another study by Munakata (2008) shows that there is no
standard definition for Artificial Intelligence, but he says according to the Webster’s new world
college dictionary “AI is the capability of computers or programs to operate in ways to mimic
human thought processes, such as reasoning and learning” (p.1). This study helps to identify the
dictionary definition of Artificial Intelligence. However, the proper definition of Artificial
Intelligence was provided in the study by Swarup (2012). He states that even though there are
many text book definitions of AI, the actual definition of AI was provided by John McCarthy,
who coined the term in 1956. John McCarthy defines AI as “the science and engineering of
making intelligent machines” (p.1). All these studies and findings are important to understand
the definition of Artificial Intelligence from different researcher’s points of view. These
researchers emphasized understanding the need and importance of the Artificial Intelligence.
Artificial Intelligence has been used for various purposes in different fields. Russell and
Norvig (2003) state that Artificial Intelligence systems should possess the following capabilities:
“Natural language processing to enable it to communicate successfully in English. Knowledge

3
representation to store what it knows or hears; automated reasoning to use the stored information
to answer questions and to draw new conclusions; machine learning to adapt to new
circumstances and to detect and extrapolate patterns. Computer vision to perceive objects, and
robotics manipulate objects and move about” (p. 3). This study helps to differentiate the
characteristics of an Artificial Intelligence system. In addition, Jackson (1985) explains the
problem solving, game playing, theorem proving, semantic information processing, and pattern
recognizing are the main purpose of the Artificial Intelligence (p. 6). These studies provided the
data about the various purposes of Artificial Intelligence in computational and logical reasoning.
These findings lead researchers to emphasize extending Artificial Intelligence to real world
applications.
Artificial Intelligence is classified into different fields based on the application.
According to Munakata (2008), Neural Networks, Fuzzy Systems, Data Mining, and Machine
Learning are the major fields of Artificial Intelligence (p.3). He defines Neural Network as
“Computational models of the brain” and Fuzzy Systems as “A technique of continuation to a
paradigm, especially for discrete disciplines such as sets and logic” (p.3). Another study shows
that neural network applications are crucial for solving practical problems related to cost, speed
of operation, reliability, ease of maintenance, and development (Introduction to Neural
Networks, 1995). This research provided the insight towards major fields of Artificial
Intelligence. In addition to these definitions, Black and Ertel (2011) provided a definition for
Data Mining and Machine Learning. They state that Data Mining is “the task of learning
machine to extract knowledge from training data” while Machine Learning is “the field of study
that gives computers the ability to learn without being explicitly programmed” (p.161, 165). In

4
brief, this research highlighted the prominent fields of Artificial Intelligence, which are used
extensively in computer technology.
The review of the literature shows the history, definition, purpose, and fields of Artificial
Intelligence. Many researchers focused on various aspects of Artificial Intelligence; however,
only a few researchers focused on solving real world issues using Artificial Intelligence. Very
few scientists focused on studying Artificial Intelligence concepts such as neural networks,
machine learning, and fuzzy systems to solve real life problems. This paper explores the
potential application of Artificial Intelligence techniques in resolving real time issues such as
developing brain inspired computing using neural networks, resolving network traffic using
machine learning, and solving student performance issues using fuzzy systems.
The first real time application of AI is developing brain inspired computing using neural
network concepts. Understanding human brains for computing requires study of the neural
networks. Munakata (2008) defines “A neural network (NN) as an abstract computer model of
the human brain” (p.7). A machine can be simulated to mimic the human brain. Beiye et.al
(2015) proposed a model to understand the pattern on how the human brain works. They
demonstrated the model with three main components: training the data, model selection and
training/testing (p.1). An artificially intelligent machine is trained with a specific set of
predefined data to select the required model. The machine is trained to verify the authentication
of the user request. For example, a user sends money and withdraws request from the bank. In
this scenario, an artificially intelligent machine is trained to understand how human requests are
made. The machine will decide a prototype model to interact with humans. Moreover, the
machine is intelligent to differentiate between a human request and a robot request or a fake
request from the hacker. Beiye et al (2015) demonstrated this using test privacy model shown

5
below. This model will help to investigate the problem of understanding the behavior and
communication patterns of the users of social networks.
Figure 1: Test privacy model (Beiye et.al, 2015, p.2)
Pattern recognition is one of the ways to train machines to behave like a human brain.
Beiye et.al(2015) conducted an experiment where machines are provided with some sample
handwritten material to analyze and recognize the pattern (p.4). For example, the figure shown
below is the sample of handwritten numbers. An artificially intelligent system incorporated
machine can recognize the pattern below and can identify different numbers. This kind of pattern
recognition is mainly incorporated to authenticate unique users. Since the handwriting of one
person varies from another person, artificially intelligent systems will easily identify the unique
users.
Figure 2: Training digit pattern samples (Beiye et.al, 2015, p.4)

6
A study by Roy, Sharad, Deliang, and Yogendra (2014) demonstrated the usage of spin
torque devices to low energy non Boolean computing. This research emphasized studying brain
neural network patterns to design non Boolean computing. According to Ishak and Siraj (2002)
“the basic element of the brain is a natural neuron” (p.4). Roy and team (2014) conducted an
experiment where neurons are studied to understand the pattern of image recognition. They
further analyzed how neuron connects and transfers data across them. According to them each
neuron transaction is associated with a voltage. Roy and Colleagues’ (2014) experiment shows
that the communication pattern is studied by varying voltage levels. Association of spin torque
devices demonstrated with high voltage the image quality is distracted, whereas with low voltage
the image is processed completely.
Figure 3: Compressed vector representation of stored image (Roy, Sharad, Deliang, & Yogendra, 2014, p.3)
Brain simulation experiments are also supported by the other research, where the neural network
pattern was analyzed to provide statistical simulation on social networking sites (Aabed &
AlRegib, 2012, p. 111). Pattern recognition and brain simulation research demonstrated the
features of brain computing to secure the confidential data from attackers. In short, neural
network concepts can be incorporated into a machine which can behave like a human brain. This
sophisticated machine can be utilized in developing applications that are useful to humans.

7
The second real time application of AI is resolving network traffic using machine
learning techniques. Machine learning knowledge is required to solve critical problems.
According to Chandra and Hareendran (2014), “Machine Learning is a field of study that gives
computers the ability to learn without being explicitly programmed” (p.161). De Souza, Matwin,
and Fernandes (2014) state that usage of the Internet has increased since 1980s (p.1). In the
Internet all computers are connected to different networks and associated with Network Address
Translations (NAT). This NAT is required to establish communicate between computers.
According to Goksen (2014), NAT is generally associated with one computer connected to
network using Local Area Network (LAN). Multiple users can connect to the LAN using Wi-Fi
where NAT can be provided by Internet Service Providers (ISPs). The ISPs will provide a unique
private Internet Protocol (IP) address to each connection. NAT encapsulates this private address
and displays the different public address to the external users in the network.
Since the number of users of the Internet has increased, resolving NAT has become a
challenge. The servers should be designed with sophisticated technology to interpret and resolve
network addresses. For example, multiple users will be accessing the same website from
different locations at the same time. If the server cannot resolve the IP address, users will not
receive the data. The server design should be robust to resolve network traffic. Goksen and
colleagues (2014) provided a solution for this network traffic problem using machine learning
techniques. They proposed an approach to identify the potential NAT devices on given traffic
traces. They proposed an algorithm called C4.5 to resolve this issue. This algorithm uses
machine learning technique to resolve IP addresses. For example, the table below represents the
data flow from multiple websites. Gokcen’s (2014) C4.5 algorithm will segregate unencrypted
and encrypted data from multiple addresses. After segregation, the algorithm will resolve the IP

8
addresses from different ISPs and process them to identify the traffic type. According to
Foroushani and Zincir-Heywood (2013), a machine incorporated with C4.5 algorithm will
identify 36000 files from different website at any given point of time (p. 75).
Table 1: Network flow from different IP addresses (Foroushani, V.A., & Zincir-Heywood, A.N., 2013, p
.75)
Along with C4.5 algorithm, Gokcen (2014) provided another Machine Learning based
approach to solve network traffic is Naïve Bayes algorithm. In this research two different data
sets were collected from two different network flows such as Nims- NAT and Partner-NAT. The
same data sets had provided to two algorithms C4.5 and Naïve Bayes for analysis. Statistics
conducted by Gokcen (2014) shows that the data was collected from HTTP and SSH networks.
His C4.5 algorithm provides solution based on data analysis, whereas Naive Bayes algorithm a
standard baseline based on statistical learning. The table below represents the results of the data
analysis report from these two algorithms. Gokcen’s C4.5 algorithm evaluates better than Naïve
Bayes for NAT data set. On the other hand, Naïve Bayes algorithm performed well for other data
set compared to C4.5 algorithm.

9
Table 2: Comparison between C4.5 and Naïve Bayes algorithms (Gokcen, 2014, p. 138)
In brief, this reducing network traffic traces experiment provided an insight into network traffic
related problems and demonstrated algorithms to resolve the traffic traces. Overall, this research
helped to resolve network traffic issues using machine learning techniques.
The third real time application of AI is solving student performance issues using fuzzy
logic. Fuzzy systems is one of the mathematical and logical approaches used by researchers to
solve the problems. According to Munakata (2008), “Fuzzy Systems is a technique of
continuation to a paradigm, especially for discrete disciplines such as sets and logic” (p.3).
Yildiz and Baba (2014) state that “The fuzzy logic concept was introduced in 1965 as a
mathematical way to represent linguistic variables” (p. 1023). They further explain that Fuzzy
logic defines problem probability between 0 and 1 like 30% “normal”, 40% “good” and %30
“bad” (p. 1023). Therefore, fuzzy system is the best approach to analyze the performance of any
system. Yildiz and Baba (2014) classified fuzzy logic with four important steps: fuzzification,
knowledge base, decision making scheme and defuzzification (p. 1023).
Fuzzy logic approach is used to solve student performance issues. Yildiz and Baba (2014)
proposed the system interface to assess the student grades using fuzzy logic (p. 1024). They
developed a fuzzy evaluation system shown below. This model collected real time data on

10
student grades periodically to evaluate the performance. This approach first determines the main
criteria for evaluation and associates each criteria with a predefined weight. Yildiz and Baba
(2014) studies demonstrated how students score is calculated using fuzzy interface. Their student
evaluation system will continuously assess the grade after peer, student, and group assessment
using fuzzy interface.
Figure 3: Student Evaluation System (Yildiz & Baba, 2014, p. 1024)
After the assessment each student is associated with weight such as poor, unsatisfactory,
average, good, and excellent based on their performance. In addition, the student evaluation
system by Yildiz and Baba (2014) supports to calculate the curve point average. However,
sometimes the fuzzy system exhibit error, but that is very minimal. Overall, fuzzy logic based
student evaluation system provides more realistic and reliable education assessment.
Application of fuzzy system is also demonstrated by Sardesai, Sambarey, Kharat,
and Deshpande (2014) by providing a case study. They utilized type 1 Fuzzy system interface to
predict patient data (p.2). This case study involves collecting data from 226 gynecology patients

11
and applying fuzzy interface system to apply eight experts perceptions based on patients data.
The case study on fuzzy logic application on gynecology by Sardesai and team (2014) involves
fuzzification of eight perceptions to create production rules followed by aggregation and
defuzzification of patient data to provide diagnosis (p.2). In brief, fuzzy logic based system is
reliable to analyze enormous runtime user data and provide valuable result evaluation of the
collected data.
In conclusion, Artificial Intelligence is the important modeling technique in
computational study. Artificial Intelligence has facilitated the resolution of many real time
issues. The future of Artificial Intelligence is promising. A considerable amount of research is
scheduled in Artificial Intelligent Robots which would replace human in the future. The neural
network based brain simulation can be extended to identify the communication pattern in social
networking sites. The Machine learning based approach to solve network traffic can be further
analyzed with different NAT behavior to explore C4.5 based classifier for automatic signature.
The fuzzy logic based approach can be extensively used in the healthcare system to manage
medical reports. This paper has explored the potential application of Artificial Intelligence
techniques in resolving real time issues such as developing brain inspired computing using
neural networks, resolving network traffic using machine learning, and solving student
performance issues using fuzzy systems.

12
References
Aabed, M.A., & AlRegib, G. (2012). Statistical modeling of social networks activities. Emerging
Signal Processing Applications (ESPA), 12, 111-114. DOI: 10.1109/ESPA.2012.6152458
Beiye, L., Chunpeng, W., Hai, L., & Yiran, C. (2015). Cloning your mind: Security challenges in
cognitive system designs and their solutions. Design Automation Conference (DAC), 15,
1-5. DOI: 10.1145/2744769.2747915
Chandra, V., & Hareendran, A.S (2014). Artificial Intelligence and Machine Learning. New
Delhi, India: PHI learning private limited.
De Souza, E.N., Matwin, S., & Fernandes, S. (2014). Traffic classification with on-line ensemble
method. Global Information Infrastructure and Networking Symposium (GIIS), 14, 1-4.
DOI: 10.1109/GIIS.2014.6934280
Ertel, W., & Black, N. (2011). Introduction to Artificial Intelligence. New York, NY: Springer.
Foroushani, V.A., & Zincir-Heywood, A.N. (2013). Investigating application behavior in
network traffic traces. Computational Intelligence for Security and Defense Applications
(CISDA), 13, 72-79. DOI: 10.1109/CISDA.2013.6595430
Gokcen, Y. (2014). Can We Identify NAT Behavior by Analyzing Traffic Flows? Security and
Privacy Workshops (SPW), 14, 132-139. DOI: 10.1109/SPW.2014.28
Introduction to Artificial Neural Networks. (1995). Electronic Technology Directions to the Year
2000, 1995. 12, 36 - 62. DOI: 10.1109/ETD.1995.403491
Ishak, W., & Siraj, F. (2002). Artificial Intelligence in Medical Applications: An Exploration.
Health Informatics Europe Journal. Retrieved from
http://citeseerx.ist.psu.edu/viewdoc/download?doi=10.1.1.102.4631&rep=rep1&type=pdf
Jackson, P. C. (1985). Introduction to Artificial Intelligence. New York, NY: Dover Publications
Munakata, T. (2008). Fundamentals of New Artificial Intelligence. New York, NY: Springer.
Roy, K., Sharad, M., Deliang. F., & Yogendra, K. (2014). Brain-inspired computing with spin
torque devices. Design, Automation and Test in Europe Conference and Exhibition
(DATE), 14, 1-6. DOI: 10.7873/DATE.2014.245
Russell, S., & Norvig, P. (2003). Artificial Intelligence: A Modern Approach. New Jersey, USA:
Pearson Education.
Sardesai, A., Sambarey, P., Kharat, V., & Deshpande, A. (2014). Fuzzy logic application in

13
gynecology: A case study. Informatics, Electronics & Vision (ICIEV), 2014 International
Conference, 14, 1-5. DOI: 10.1109/ICIEV.2014.6850715
Swarup, P. (2012). Artificial Intelligence. International Journal of Computing and Corporate
Research. Retrieved from http://www.ijccr.com/july2012/4.pdf
Yildiz, Z., & Baba, A.F. (2014). Evaluation of student performance in laboratory applications
using fuzzy decision support system model. Global Engineering Education Conference
(EDUCON), 14, 1023-1027. DOI: 10.1109/EDUCON.2014.6826230

Applications of AI

More Related Content

What's hot (16)

Viewers also liked (17)

Similar to Applications of AI (20)

Applications of AI