European Journal of Engineering and Technology Research

ISSN: 2736-576X

Machine Learning in Pattern Recognition

Chetanpal Singh

classes for every item of the information on the other hand in

Abstract — Supervised or unsupervised classification is the unsupervised the classes are attached with the help of the
main objective of pattern recognition. The statistical approach likeness of classes, as well as the input pattern, is also allowed
is the most popular approach that is practised among the several
[4].
frameworks where pattern recognition is initially formulated. In
the recent past, the neural network technique and the Transfer Learning, Multi-instance Learning, as well as the
methodology scheme from the statistical learning theory have various emerging trends in the deep learning technique, are
garnered the attention of people. It requires proper attention to some of the new learning methods that have been utilized for
deal with the design of the recognition system. There are several the solution to combat pattern recognition issues. Some of the
issues associated with the design of the recognition system. They common pattern recognition problems are visual tracking,
are the pattern class definition, sensing environment and
text classification, object detection, drug activity prediction,
representation extraction and selection of features, cluster
analysis, classifier design, learning, and choosing the training and image classification [5]. To predict the radiation of solar,
and test samples. There is no solution to the general issue of the machine learning technique has been beneficial by
recognizing complex patterns associated with arbitrary integrating with artificial neural networks. Machine learning
patterns. Data mining, web searching, and retrieval of techniques combined with artificial neural networks used
multimedia are the various emerging applications that require frequently are SVM, Random Forest, Support Vector
proper and effective regulation techniques. The main purpose of
Regression (SVR), Regression Tree, k mean and Boosting
this paper is to give a detailed overview of the various methods
that can be used in the different stages of the pattern recognition methodologies [6]. The performance of the Convolutional
system. The paper also aims to figure out the research topics in Neural Network to determine nodule detection in the medical
the application that can be highlighted in this challenging field. field for analyzing tasks can be enhanced by introducing the
new imaging approach known as Massive Training Artificial
Keywords — Classification, Neural Network, Machine Neural Networks to eliminate the various handcrafted
Learning, Pattern Recognition, Pattern Matching, Security. features [7]. The proposed model can identify and classify the
focal defect without the usage of any high-level semantic
feature.
I. INTRODUCTION
Moreover, face recognition, handwritten recognition, and
There are many systems that data mining includes for speech recognition, fingerprint recognition are some of the
preprocessing, analyzing, and interpreting the data. There are popular pattern recognition solutions to the various biometric
two fields in which these strategies can be categorized: methods as the iris images are very much distinctive and
pattern recognition and machine learning. The main purpose stable when the age of the person is taken into account [8].
of pattern recognition is recognizable proof of the various The design and development of a system have a lot to do
verifiable things and relationships. In other words, pattern with pattern recognition that tends to recognize patterns
recognition plays a crucial role in the extraction of patterns within the data. The main objective of the pattern recognition
from the data that is given [1]. The pattern recognition program is to identify a scene in the real world and describe
process is related to image arises even though this is not the scene that will be beneficial for achieving some task [9].
regarded as the main sort of application. Sensor and pattern innovation plays a crucial role in gathering
The machine learning technology deals with extraction of real-world observation. Numeric symbol information is
the generalized knowledge from information that can be computed by the feature extraction mechanism from the
utilized for the prescient errand. To combat and defeat the observation. Once numeral symbol information is computed
pattern recognition issues, there are a few arrangements that these features that are accepted are classified with the help of
have been proposed [2]. The paper has highlighted the a classified. There are many procedures to ensure the efficient
machine learning strategies among the arrangements that description of the patterns in the process that has been used
have been proposed to defeat the pattern recognition problem. by pattern recognition [1]. Hence it makes sense to include
Rule-based learning, decision trees, support vector machine, knowledge techniques for dealing with pattern evolution for
and Naive Bayes classifier that has been incorporated into the enhancing the recognition rate as well as accuracy. To
strategy. The methodologies that have been proposed ever determine the better pattern recognition method, there has
have the same principle that doesn't need any determination been a review paper designed that analyzes the machine
of rules expressly [3]. The characteristics of machine learning learning related to pattern recognition with the help of various
can be done by the input data that is supervised and recent technologies to enhance the performance accuracy of
unsupervised classification. The directed classification the potential solution for the various issues surrounding data
approach where classes are determined by the pre-stored mining and other related fields [10].
learning information combined with pre-characterized

Submitted on March 15, 2023. C. Singh, Dr., Holmes Institute Melbourne, Australia.
Published on April 11, 2023. (e-mail: Csingh@ holmes.edu.au)

DOI: http://dx.doi.org/10.24018/ejeng.2023.8.2.3025 Vol 8 | Issue 2 | April 2023 63

 European Journal of Engineering and Technology Research
ISSN: 2736-576X

Fig. 1. Pattern Recognition in Machine Learning [11].

II. PATTERN RECOGNITION Yan et al. [13] has come up with a face recognition system
Pattern recognition is a part of a machine learning that is based on a one-class support vector machine required
algorithm that is widely used to figure out patterns. The for mobile devices that run on the Symbian operating system.
pattern recognition approach classified data based on Nokia 6680 mobile is tested for evolution and the outcome
statistical information derived from patterns and has shown that EER of 7.92 % and 3.95% per cent can be
representations [11]. The labeled training data is utilized in achieved based on the global threshold and the individual
this technique to deal with pattern recognition systems. The threshold respectively.
specific input value is attached with the level that in turn Monté-Rubio et al. [14] demonstrated the interest in
produces a pattern with output. When there are no label data implementing pattern recognition methods to anatomical
other computer algorithms are utilized to identify unknown neuroimaging data. However, there has been a very little
patterns [2]. investigation done about ways to derive image features to
make the most proper prediction. The Gaussian process
A. Research Objective machine learning approach was utilized in this work for
This research papers aims at discussing various predicting the age, gender, and body mass index of the
applications of machine learning techniques in pattern various objects in the data set.
recognition. The paper will discuss the academic Silasai and Khowfa [15] came up with a low-cost biometric
contributions of various scholastics in the literature review authentication system for mobile phones that can be used for
section along with an in-depth understanding of the benefits face detection, registration illumination, and normalization
of pattern recognition, the challenges faced during its verification as well as information fusion. The Anal error rate
implementation and the importance of machine learning tools of 2% in the analysis can be achieved with the help of the
for pattern matching and recognition. system.
Findling et al. [16] proposed shot face unlocks which is a
B. Research Motivation
popular mobile device unlock mechanism that uses all the
Pattern recognition assists in the field of computer vision, information that is there from a 180-degree pan shot of the
content, record examination, data processing, speech mobile device around the head of the user. For the face
recognition, text classification, image processing as well as recognition approach the evolution of different support vector
neural network systems [10]. Thus, pattern recognition has a machines, as well as neural networks, was done, and the result
major role in machine learning and there are various indicates the feasibility of the approach.
applications of it when it is combined with machine learning Nair et al. [17] has come up with an approach of
like in the application of biometrics, bioinformatics, big data determining face authentication schemes with the help of a
analysis, and developed data science. Pattern recognition is Haar-like feature using a boost for the sake of face and eye
an approach to grouping input information in items as well as detection. The result that the approach has obtained is very
classes according to the key highlight [1]. One upon reading satisfying as it indicated the feasibility of face authentication
the paper will get to know about the various analyses done on on mobile devices. The proposed result achieved an average
pattern recognition methodologies with the help of machine authentication rate of 82 % for the small-sized pieces and the
learning techniques. rate was as high as 96% for medium size devices.
Xi et al. [18] have proposed a correlation related to the face
authentication scheme for the mobile phones that could help
III. LITERATURE REVIEW analyze the relationship of cross-correlation output peaks
A. Reviews on Pattern Recognition Application arising from the selected subdivision region. The scheme was
Sensor assisted facial authentication method was proposed implemented on the Nokia S60 CLDC emulator with the help
by Chen et al. [12] where the implementation of motion and of Java ME.
light sensor 2 to combat with 2D Media attain and virtual B. Algorithm in Pattern Recognition
camera attack boosted the results of the experiences Below are the various algorithms that have been used in
showcased a good detection rate of 95-97 per cent and false pattern recognition.
alarm rate of 2-3 per cent based on 450 trials in the real
settings.

DOI: http://dx.doi.org/10.24018/ejeng.2023.8.2.3025 Vol 8 | Issue 2 | April 2023 64

 European Journal of Engineering and Technology Research
ISSN: 2736-576X

6) Neural network-based algorithm

Hybrid The parallel structures that are known as neurons are
considered by this model. This is one of the most suitable
models for pattern recognition because it features superior
Neural
Network-based learning abilities. The feed-forward back propagation neural
Statistical
network is the best example of a neural network that has been
used for pattern recognition.
Pattern C. Tools used for Pattern Recognition in Machine
Recognition
Algorithms
Learning
1) IBM Watson Studio
The studio is provided by IBM and is an open-source tool
Fuzzy-based Structural that plays a crucial role in data analysis as well as machine
learning. The tool is effective for building and deploying
machine learning models on the desktop [18].
Template 2) Amazon lexes
Matching
This tool is provided by Amazon and is open-source
software that plays a crucial role in building intelligent
Fig. 2. Pattern Recognition Algorithms (Self-Created) [19].
conversion agents like chatbots using text and speech
recognition [21].
1) Structural algorithm
3) R studio
When the pattern recognition process is not simple, the
structural algorithm plays a crucial role, and the algorithm is This tool uses the R programming language. The studio is
imperative at the time of multidimensional entities usage. an integrated development ecosystem for creating and testing
Patterns can be segregated into subclasses hence it forms a the pattern recognition model [4].
hierarchical structure [19]. The relationship between the 4) Google cloud autoML
elements in the system is defined by the structural model. Google cloud autoML is used to build high-quality
2) Hybrid algorithm machine learning models this technology plays a crucial role.
This type of algorithm is used to make hybrid models This technology uses a neural network and integrates learning
where one can see multiple classifiers to identify patterns. as a base for model construction.
The classifiers in the hybrid algorithm experience training 5) Microsoft Azure machine learning studio
related to features basis [8]. There are a set of combiners and The tool is provided by Microsoft and is used to drag and
classifiers that are utilized to conclude it. The decision drop concepts for building as well as deploying the machine
function is also used to judge the accuracy of the classifier. learning model [19]. The graphic user interface is offered by
3) Statistical algorithm this tool by looking at the environment for model construction
The statistical algorithm is used in an organization to as well as usage.
develop a statistical model. This model has patterns that are D. Scope of Pattern Recognition and Machine Learning
described based on the features. With the help of this model, 1) Image processing
predictions can be done of the probabilistic nature of the
pattern [20]. The various features that were chosen are The two types of image processing are digital image
utilized to make a cluster. The system accordingly adapts by processing and analog processing. Digital image processing
analyzing the probability distribution of the pattern. Once this refers to the use of intelligent machine learning algorithms to
is done the pattern experience for the processing and the enhance the image quality that is obtained from the source
model are subjected to distinct patterns to figure it out. that is kept at a distance such as the satellite [5].
4) Template matching algorithm 2) Recommender system
The template matching algorithm plays a crucial role to Recommender systems are popular among websites that
develop a template-matching model that features a simple are committed to online shopping. Recommender systems
recognition model. The images are used by this model to collect data based on the customer purchase and suggest it
determine the similarity and the matched pattern from the with the help of machine learning algorithms by figuring out
pattern recognition model is stored as a template [4]. the trends in the pattern of the purchase of the customer [4].
However, the major disadvantage of this model is that it is not 3) Analysis
so efficient when it comes to recognizing distorted pattern. To identify important data, trend pattern recognition plays
5) Fuzzy based algorithm a crucial role. These kinds of trends are also helpful in making
In this algorithm fuzzy logic is applied that uses the truth future predictions [11]. Everyone requires analysis whether it
values from 0 and 1. In this model, some rules are applied so is technical or nontechnical.
that the given input can be matched with the output [13]. The
results produced by this fuzzy-based algorithm are quite
satisfactory because it is competent for uncertain domains.

DOI: http://dx.doi.org/10.24018/ejeng.2023.8.2.3025 Vol 8 | Issue 2 | April 2023 65

 European Journal of Engineering and Technology Research
ISSN: 2736-576X

4) Data mining
Mining is a process of extracting useful information from
huge amounts of data where the source is heterogeneous.
Data mining helps in obtaining meaningful data and is used
for predicting and data analysis [2].
5) Bioinformatics
Bioinformatics is known to be a branch of science where
computational tools and software are used for predictions
based on biological data. A good example of bioinformatics
is when someone has discovered a new protein in the
laboratory, but the protein sequence is unknown [16]. With Fig. 4. Phases in Pattern Recognition [10].
the help of a bioinformatics tool, one can compare the
unknown protein with the large number of proteins that are
stored in the database. This will help in editing the sequence V. RESULTS AND DISCUSSION
that is related to two similar patterns.
A. Results
There is no doubt about the growing popularity of machine
IV. METHODS AND MATERIALS learning, and it cannot be avoided as we implement it many
times even without our knowledge. The supervised or
A. How does Pattern Recognition work?
unsupervised are the ways by which machine learning
The concept of learning is used to do pattern recognition. techniques are frequently sorted [10]. Talking about
To provide accurate results, the concept of learning plays a supervised methodology, this algorithm expects a person to
crucial role in assisting pattern recognition systems to train provide information as well as wanted yield. Once the
and adapt to providing the result [15]. preparation is finished, the calculation will be done to ensure
To train the system the data set section plays a crucial role, what was the result that came out from the new information
and the rest element is used for testing. Listed below are the [1]. Whereas, in the case of unsupervised learning, there is no
images that highlight the usage of data for training as well as way that the algorithm will be prepared with the required
testing. result information and the unsupervised methodology uses an
The images of the data are contained by the training set that iterative approach that is known as deep learning to audit
plays a significant role in training the model. To come up with information and come up with conclusions [9]. The
criteria for the decision output the training rules are utilized. unpredictable tasks are done using unsupervised learning
Moreover, to match the given input data with the decision the methods and the supervised learning framework has no role
training algorithm also plays a significant role [2]. The next to play in doing the unpredictable task.
step is followed by facilitating training where the algorithms The research paper has highlighted the recent pattern
and rules are implemented. The information is used by the recommendation metallurgy using different solutions for the
system and collected from the data to come up with results. various classification problems which have been
The accuracy of the system is validated by the testing set demonstrated as well as reviewed by focusing on the
and the testing data also has a major role to play in checking techniques related to machine learning [6]. Moreover, upon
whether the output is accurate or not [1]. reading the newspaper will get a detailed analysis of the
There are five main phases in which the pattern recognition pattern recognition that includes face recognition and
system work. handwritten cursive word recognition, speech signal
Listed below are the various phases explained: recognition as well as iris recognition.
1. Sensing: the input data is converted into analog data in The review has concluded that the invariant pattern
this phase in the pattern recognition system. regulation can be implemented in many applications like
2. Segmentation: the segmentation phase makes sure that character as well as face recognition [3]. On the other hand,
the object that is sensed ar isolated. according to our research that was done early in statistical
3. Feature extraction: extraction phase recognizes the pattern recognition, the emphasis on the instruction of the
property of the object and then send it for the next imperial feature was not an easy task. In the recent past, there
classification [21]. has been an activity in shaping invariant recognition methods
4. Classification: the next phase is the classification phase that do not need any invariant feature like the nearest
where the object that is sensed is categorized into a neighbour classifier with the help of tangent distance and
group. deformable template matching [15].
5. Post-processing: in the post-processing phase further
considerations are made before the final decision is B. Advantage of Pattern Recognition
taken [5]. There are a lot of techniques that provide a lot of benefits
to an individual. This not only helps in the analysis of trends
but also in making future predictions [2].
• The identification of objects is easily done at varying
distance and angle,
• It becomes easy and highly automated,

DOI: http://dx.doi.org/10.24018/ejeng.2023.8.2.3025 Vol 8 | Issue 2 | April 2023 66

 European Journal of Engineering and Technology Research
ISSN: 2736-576X

• Pattern recognition is not rocket science and it does not Data storage: you should be sure to have ample space for
need any out of the box thinking ability [11], data storage so that the information can be properly
• Pattern recognition is very useful in the field of the processed.
finance industry in making valuable predictions related Data quality: the incoming data for the algorithm as well
to the sale, as the training sets is important because it comes from reputed
• Pattern recognition is a comprehensive solution to the sources and does not have any noise that is irrelevant
real-time problem [19], information for the decision-making process [18].
• It is very helpful in the medical field of forensic Neural network opacity: one of the best instruments for
analysis. having business insights is Neural network opacity.
Pattern recognition helps in getting business insights, but
C. Importance of Pattern Recognition in the Field of you should be careful to account for the neural network
Machine Learning opacity [19]. That means sometimes you might fail to explain
• The pattern recognition figures out and analyze the the output and what you can do with them. The outcome is
smallest or the untraceable data, combining the multitude of neurons within the ecosystem.
• The classification of unseen data can be done,
• The object can be recognized easily that is at varying
distance,
• It plays a significant role in making proper decisions.
D. Application of Pattern Recognition
1) Trend analysis
Pattern recognition plays a significant role in learning the
trend in the given data where the analysis can be done [20].
An example of trend analysis is identifying the recent trends
in the field of sales that are made by a specific company or
organization.
2) Assistance
There is no doubt in the fact that pattern is a very important
part of our daily life as it provides a lot of help in day-to-day
activities [2].
There is a lot of software in the market that utilizes
machine learning algorithms in making predictions related to
the presence of obstacles and notifying the user about any
mishaps.
3) E-commerce
Fig. 5. Applications of Pattern Recognition [15].
Appropriate results are provided by the visual search
engines by recognizing the proper item related to its
specification. Many websites are specifically related to online VI. CONCLUSION
shopping that uses the recommender system [7].
Machine learning is becoming popular in the 21st century
4) Computer vision and the reason why it has got a lot of demands is because of
The user makes interaction with the computer system by the application advantage. Machine learning Technology will
providing an image or video as the input and the machine then bring revolution in the industry because of its amazing
differentiates it with millions of images that are there in the capabilities and there are a lot of scopes of meeting machine
database to identify similar patterns [18]. learning in various fields which include pattern recognition,
5) Biometric devices data mining, analysis, etc. The pattern recognition link with
Biometric devices help in securing authentication as well machine learning is used in our day-to-day life whether it is
as security by using face recognition and fingerprint detection technical or nontechnical.
Technologies [8]. A. Future Work
E. Machine Learning and Pattern Recognition Challenges In the recent past, the deep neural network has a lot of roles
The machine learning and pattern recognition process is in pattern recognition and machine learning. Hence one can
very simple. However one has to care about a lot of things conclude from this review that there is the proper direction of
before implementing pattern recognition into business. fruitful research in the field of pattern recognition like face
Data processing power: if your business is related to recognition, speech and signal recognition, fingerprint
patients then there is no need to worry about the data recognition as well as iris recognition.
processing power however if you are surrounded by big data B. Open Research Questions
projects, then you should be concerned with your Pattern recognition is a popular one that offers the
infrastructure [20]. recognition of the data regulations as well as pattern
recognition. The application that accepts this process has

DOI: http://dx.doi.org/10.24018/ejeng.2023.8.2.3025 Vol 8 | Issue 2 | April 2023 67

 European Journal of Engineering and Technology Research
ISSN: 2736-576X

experienced significant growth over the recent past. This correlation-based face authentication scheme and its java me
application. Concurr Comp Pract Exp, 2012; 24: 1066-1085.
application has found a solution to various real-life [19] Zhu X, Wang Z, Lin P, Ma Z, Yu Z. Algorithm and Technology
challenges with the usage of training data, testing data, and Application of Image Recognition Based on Artificial Intelligence.
classifiers. Journal of Physics: Conference Series, 2021; 2136(012062): 1-6.
[20] Jangapally T, Hiwarkar T. Performance Analysis of Pattern
Moreover, pattern recognition can come up with a more
Recognition Algorithms Using Artificial Neural Networks.
intelligent process that assists in various digital technologies. International Research Journal of Modernization in Engineering
The technique has the potential to help in the advancement of Technology and Science, 2020; 2(7): 1501-1508.
robotics and automation. [21] Veena S, Shankari T, Sowmiya S, Varsha M. A Survey on Tools Used
For Machine Learning. International Journal of Engineering Applied
The concept of pattern recognition can also be used in Sciences and Technology, 2020; 4(9): 116-119.
autonomous cars. The autonomous car is the next big thing
that is gaining momentum and the concept of pattern
recognition in this field might increase because autonomous
cars are where there is the necessity to detect objects, cars,
people as well as traffic lights.

REFERENCES
[1] Bhamare DP, Suryawanshi P. Review on Reliable Pattern Recognition
with Machine Learning Techniques. Fuzzy Information and
Engineering, 2018; 10(3): 362–377.
https://doi.org/10.1080/16168658.2019.1611030.
[2] Armengol E, Boixader D, Grimaldo,F. Special Issue on Pattern
Recognition Techniques in Data Mining. Pattern Recognition Letters,
2017; 93: 1–2. https://doi.org/10.1016/j.patrec.2017.02.014
[3] Kumar S, Gao X, Welch I. A machine learning based web spam
filtering approach. In 2016 IEEE 30th International Conference on
Advanced Information Networking and Applications (AINA), 2016.
[4] Ermushev S, Balashov A. A Complex Machine Learning Technique for
Ground Target Detection and Classification. Int J Appl Eng Res., 2017;
11(1): 158–161.
[5] Wu J, Yu Y, Huang C, Yu K. Deep multiple instance learning for image
classification and auto-annotation. Computer Vision and Pattern
Recognition, 2015. https://doi.org/10.1109/cvpr.2015.7298968.
[6] Notton VG, Kalogirou S. Machine learning methods for solar radiation
forecasting: a review. Renew Energ., 2017; 105: 569–582.
[7] Tajbakhsh N, Suzuki K. Comparing two classes of end-to-end
machine-learning models in lung nodule detection and classification:
MTANNs vs. CNNs. Pattern Recognit., 2017; 63: 476–486.
[8] Aginako N, Echegaray G, Martínez-Otzeta J. Iris matching by means
of machine learning paradigms: a new approach to dissimilarity
computation. Pattern Recognit Lett., 2017; 91: 60–64.
[9] Saii MM. Classification of Pattern Recognition Techniques Used Deep
Learning and Machine Learning. International Journal of Computer
Science Trends and Technology (IJCST), 2019; 7(3): 165-173.
[10] Omarov B, Cho YI. Machine learning based pattern recognition and
classification framework development. In 2017 17th International
Conference on Control, Automation and Systems (ICCAS 2017).
Ramada Plaza, Jeju, Korea, 2017.
[11] Ushmani A. Machine Learning Pattern Matching. International
Journal of Computer Science Trends and Technology (IJCST), 2019;
7(2): 4-7.
[12] Chen S, Pande A, Mohapatra P. Sensor-assisted facial recognition: an
enhanced biometric authentication system for smartphones. In
Proceedings of the 12th Annual International Conference on Mobile
Systems, Applications, and Services, 2014.
[13] Yan Z, Zhan Y, Peng Z, Liao S, Shinagawa Y, Zhang S, Metaxas DN,
Zhou X. Multi-Instance Deep Learning: Discover Discriminative Local
Anatomies for Bodypart Recognition. IEEE Transactions on Medical
Imaging, 2016; 35(5): 1332–1343.
https://doi.org/10.1109/tmi.2016.2524985.
[14] Monté-Rubio G, Falcón C, Pomarol-Clotet E. A comparison of various
MRI feature types for characterizing whole brain anatomical
differences using linear pattern recognition methods. NeuroImage,
2018; 178: 753-768.
[15] Silasai O, Khowfa W. The Study on Using Biometric Authentication
on Mobile Device. NU. International Journal of Science, 2020; 17(1):
90-110.
[16] Findling R, Hölzl M, Mayrhofer RM. Mobile match-on-card
authentication using offline-simplified models with gait and face
biometrics. IEEE Trans Mob Comput., 2018; 17(11): 2578-2590.
[17] Nair HH, Amte GS, Todase NB, Dandekar PR. Face detection and
recognition in smartphones. International Journal of Advance
Research and Development, 2018; 3(4): 177-182.
[18] Xi K, Hu J, Han F. Mobile device access control: an improved

DOI: http://dx.doi.org/10.24018/ejeng.2023.8.2.3025 Vol 8 | Issue 2 | April 2023 68