9 III March 2021

https://doi.org/10.22214/ijraset.2021.33316
 International Journal for Research in Applied Science & Engineering Technology (IJRASET)
ISSN: 2321-9653; IC Value: 45.98; SJ Impact Factor: 7.429
Volume 9 Issue III Mar 2021- Available at www.ijraset.com

A Review of Grey Scale Normalization in Machine

Learning and Artificial Intelligence for
Bioinformatics using Convolution Neural Networks
Divya Kothari1, Ajay Kumar Sharma2, Mayank Patel3
1
Research. Scholar, 2, 3Associate Professor, Department of Computer Science & Engineering, GITS, Udaipur

Abstract: These days, machine learning is a modern trending area and is an artificial intelligence technology. To make
computers function in a certain way without being specifically programmed, machine learning uses certain statistical
algorithms. The algorithms obtain an input value and by using some statistical techniques, forecast an output for this. Machine
learning's primary goal is to create intelligent machines that can think and function like human beings. Machine learning
systems can perform sophisticated processes by gathering and analyzing data, rather than adopting pre-programmed rules, by
allowing computers to perform complex jobs smartly. Exciting developments in machine learning have been seen in recent years,
which have expanded its abilities across a variety of applications. Growing data availability has helped machine learning systems
to be developed on a broad pool of examples, while the computational capabilities of these systems have been supported by
increasing computer processing power.
Keywords: Machine learning, Deep learning, Bioinformatics, Convolution neural network

I. INTRODUCTION
Machine learning helps machines to replicate and manipulate human-like behavior. Using machine learning, each interaction, each
action carried out, becomes knowledge that the computer can learn and use as training for the next occasion [1]. And there have
been computational developments within the discipline itself, which have provided significant strength to machine learning. As a
consequence of these developments,devices that operated at considerably below-human levels just a few years ago now can perform
better than humans in certain particular tasks.A report [2] says Machine learning may indeed be a key enabler for a variety of
scientific fields, advancing forward the parameters of science by processing the enormous amounts of data now being produced in
fields such as life sciences, particle physics, astronomy, social sciences, and more. For researchers to examine these huge datasets,
recognizing previously unforeseen trends or extracting unusual observations, machine learning may become a key tool.
In this paper, we address the analysis state of machine learning, discuss what it is, why it has succeeded in improving deep
rooted approaches of conventional neural networks, and, most significantly, how you can attempt to incorporate deep learning
into research activities to tackle both new and existing problems and develop improved, wiser user devices and applications. This
research is an analysis of this type of data processing that allows computers to learn and do what is intuitive to humans, i.e. to learn
from the past. It includes the basics of machine learning that answers what, how and why, the meaning, applications and
implementations. To recognise and check its potential in bioinformatics, the tech framework of machine learning is addressed. The
purpose of this study is to provide perspective into why machine learning seems to be the potential.Since it's an emerging new
technology and this technology is not known to most people. We bring forward some important postulates of this conception with
our research work.

II. MACHINE LEARNING

Machine learning is a branch of artificial intelligence that gives systems the ability to learn automatically without being specifically
programmed or without human interference, and to develop themselves from experience. The main purpose is to allow computers
learn from experience automatically.
The system will learn to differentiate between the pattern and make a relatively good prediction due to the large amount of data
produced by computers, sensors and social media users [3]. The learning factor also produces systems that can be flexible and, once
they have been implemented, continue to enhance the precision of their outcomes [4]. In 2015, for instance, researchers developed a
machine learning system that, in a limited range of vision-related tasks, exceeded human capacities on identifying individual
handwritten digits [5].

©IJRASET: All Rights are Reserved 1306

 International Journal for Research in Applied Science & Engineering Technology (IJRASET)
ISSN: 2321-9653; IC Value: 45.98; SJ Impact Factor: 7.429
Volume 9 Issue III Mar 2021- Available at www.ijraset.com

A. Requirements for Building good Systems for Machine Learning

1) Data: To predict the performance, input data is required.
2) Algorithms: Machine Learning to determine data patterns relies on some statistical algorithms.
3) Automation: It is the ability to automatically make devices run.
4) Iteration: The entire process is iterative, i.e. process repetition.
5) Scalability: In size and scale, the power of the computer can be increased or decreased.
6) Modeling: The models are generated by the modelling process according to demand.

B. Machine Learning Techniques are Grouped into Several Groups. [1]

1) Supervised Learning: The machine is provided with input and output along with feedback during the training in this process.
The computer's accuracy of predictions during training is also evaluated. The primary objective of this training is to teach
computers how to map input to output.
2) Unsupervised Learning: In this case, no such training is provided, leaving computers to find the output on their own.
Transactional data is often applied to unsupervised learning. It is used in more difficult assignments. To arrive at any
conclusions, it utilises another iteration technique known as deep learning.
3) Reinforcement Learning: Three components are used in this type of learning: agent, environment, behaviour. An agent is the
one who perceives the world, the one with whom an agent communicates and behaves in that environment. In reinforcement
learning, the primary purpose is to find the best policy possible.

C. Machine Learning Functioning

Machine learning makes use of methods equivalent to data mining processes. In terms of target function (f), mapping the input
variable (x) to an output variable, machine learning algorithms are defined (y).

It is possible to represent this as:

Y=f (x)

There is also an e error that is distinct from the x input variable. Thus the equation's more generalised form is:
Y=f(x) + e, y=f(x)

The mapping from x to y for predictions is performed on the computer. To make the most reliable predictions, this methodology is
known as predictive modeling. This function has different assumptions.

D. It Operates on the 3 Criteria that Follow

Finding flaws in algorithms for machine learning.
Develop methods for testing these possible vulnerabilities.
Implementing these preventive measures with a view to improving algorithm security.

E. Machine Learning Advantages

1) Prediction: Decision-making is quicker. By prioritising repetitive decision-making processes, machine learning offers the best
possible results. These methods construct a database of training samples and evaluate the results with other cases in the
database compared to fresh data is provided as input using a similarity measure to find the closest match and make the
judgement [6].
2) Adaptability: Machine learning offers the ability to quickly adapt to modern, evolving environment conditions
3) Innovation: Machine learning uses sophisticated algorithms that enhance the overall capacity for decision making. This helps to
build new services and models for companies.
4) Insight: Machine learning helps to understand unique data patterns and what particular steps can be taken based on them.
5) Business Development: The overall business process and workflow of machine learning would be quicker and therefore lead to
the overall growth and acceleration of the business.
6) Accurate Results: With machine learning, the accuracy of the result will be increased with less probability of error.

©IJRASET: All Rights are Reserved 1307

 International Journal for Research in Applied Science & Engineering Technology (IJRASET)
ISSN: 2321-9653; IC Value: 45.98; SJ Impact Factor: 7.429
Volume 9 Issue III Mar 2021- Available at www.ijraset.com

F. Software for Machine Learning

IBM’s ‘Watson’ uses machine learning in various ways. One of these is natural language processing – the form of machine learning
which allows computers to process written or verbal information – which Watson uses to extract information from the vast
collection of published research papers and case reports, and use this information to recommend treatment options [7].
1) Tensor Flow: It is a library of open source applications for machine learning by Google. [2] TensorFlow, along with
documentation, tutorials and other support tools, offers a library of numerical computations.
2) Amazon Web Services: Amazon has released developer toolkits along with applications ranging from interpretation of images
to facial recognition. AlsoAmazon is developing the use of delivery drones, with the first successful delivery taking place in
December 2016 [9].
3) Caffe: It is a platform for deep learning and is used in the field of voice, vision and expression in different industrial
applications.
4) Veles: This is another deep learning framework written in the language of C++ and uses python to communicate between the
nodes.
III. DEEP LEARNING
Deep Learning is part of the wider field of machine learning and is focused on learning about data representation. It is based on
artificial neural network interpretation. The algorithm for Deep Learning uses multiple computing layers. The output of the previous
layer is used by each layer as an input to itself. An algorithm or an unsupervised algorithm may be supervised by the algorithm used.
Deep Learning is specifically created to manage complex input and output mappings. It is basically designed to work like human
brain. It is another hot subject along with machine learning for the M.Tech thesis and project. One of said most popular deep
learning algorithms are: [8]
Convolutional Neural Network (CNN)
In the last few years, we have witnessed an exponential growth in research activity into the advanced training of convolutional
neural networks (CNNs) [10].

A. Deep Learning Advantages

Eliminates unnecessary costs by identifying flaws and failures in the system.
Deep learning helps to recognise defects that have been left untraceable in the system. It detects defects that are otherwise difficult
to detect.
Deep learning can inspect unusual shapes and patterns that are difficult for machine learning to identify. Deep learning can inspect
irregular shapes and patterns.
It is good at pattern recognition problems and is data-driven [11].

IV. BIOINFORMATICS MACHINE LEARNING

The word bioinformatics is a mixture of two bio-informatics concepts. Bio means biology-related, and informatics means records.
Thus, bioinformatics is a field that uses a quantitative and statistical approach to deal with the analysis and understanding of
biological data. As biological data grows exponentially, it is important to pay attention to the efficient storage and information
management, as well as to extract useful information from that data. In order to turn this diverse data into meaningful information,
suitable analytical techniques must also be used. These analytical tools and methods, or you can say machine learning tools, allow
more detailed data to be grasped and provide knowledge in terms of testable models by which we can obtain system predictions.
There are several biological areas in which machine learning tools can be used to retrieve information, followed by neural network
applications in bioinformatics;

1) In the recognition of gene area coding

2) Problems in the recognition of genes
3) Recognition and analysis of signals generated from regulatory sites
4) Detecting sequence, grouping, and characteristics
5) Genetic and Genomic Data Expression
6) Processing an image and signal

Machine Learning find its application in the following subfields of bioinformatics: Genomics, Proteomics, Microarrays, System
Biology and Text mining.

©IJRASET: All Rights are Reserved 1308

 International Journal for Research in Applied Science & Engineering Technology (IJRASET)
ISSN: 2321-9653; IC Value: 45.98; SJ Impact Factor: 7.429
Volume 9 Issue III Mar 2021- Available at www.ijraset.com

V. LITERATURE REVIEW
We observe that supervised learning offers broad opportunities in machine learning in the perspective of training and testing
machines and is quite intriguing topic for further research. Supervised learning is further categorized into numerous topics in the
viewpoint of research. Some of them are listed in table 1.

A. Regression and Classification are Broad Categories of Supervised Learning

In regression a single output value is generated in regression using training data. This value is a predictive interpretation that, after
evaluating the strength of correlation between the input variables, is calculated. Regression, for instance, can help to predict the cost
of a property based on its location, size, etc.
Classification allows the data to be divided into groups. You may use classification to decide whether or not a person is a loan
defaulter if you are thinking about extending credit to an individual. When the supervised learning algorithm labels input data into
two separate groups it is called binary classification. Multiple classifications suggest the classification of information into more than
two groups.

Table 1 : Potential fields of supervised learning that can be improvised in training and efficiency.
Supervised Area of Potential fields that can have improvised
learning research training and increased efficiency
technique
Classification Bioinformatics Use of deep learning to train models to give
better results.[12]
Regression Speech Security systems that use voice recognition to
recognisation grant customers access to their accounts.[13]
Classification Banking Robot bank-tellers that use machine learning to
respond to customer queries.[14]
Classification Image Human diseases can be detected through various
recognisation medical imaging techniques like MRI[15]

Classification Language Natural language processing techniques to

processing analyze text and detect inappropriate statements
which are indicative of phishing attacks[16]

Classification Medical To predict how well patients will respond to

different drugs used in treating depression[17]
Machine learning algorithms can help address
Regression Finance monetary policy-making [18]
Regression Weather To optimise the temperature requirements of
prediction places by predicting temperatures like Google
DeepMind [19]

VI. CONCLUSION
This paper describes a description of the mechanism of machine learning. The different machine learning algorithms based on
techniques of machine learning are also described. Examples of applications for machine learning and the requirements are
presented. Machine learning is currently automating repetitive technological tasks in many areas, but machine learning applications
in these aspects are diversifying, from machines offering legal advice to health tracking medical apps using machine learning.This
research presented the findings of study to investigate the most well-known machine learning and deep learning methods used for
classification. For organising the data, classification is very necessary so it can be easily accessed. In several different application
fields, from banking to medicine, from business to bioinformatics, these methods have gained a lot of significance.

©IJRASET: All Rights are Reserved 1309

 International Journal for Research in Applied Science & Engineering Technology (IJRASET)
ISSN: 2321-9653; IC Value: 45.98; SJ Impact Factor: 7.429
Volume 9 Issue III Mar 2021- Available at www.ijraset.com

VII. FUTURE WORK

We propose two potential categories by this review in which further research work is attainable.

A. Enhancing performance of Convolution neural network can be achieved by perform a grayscale normalization to reduce the
effect of illumination's differences. Thus, making CNN work faster. Keras model needs an extra dimension in the end which
correspond to channels. Our images are gray scaled so it use only one channel.
B. Implementation can be done for Gene Expression analysis in which acute myeloid leukemia patients can be classified into
classes to help diagnose the disease with two datasets containing the initial (training samples) and independent (test samples)
datasets.
REFERENCES
[1] JafarAlzubi, AnandNayyar and Akshi Kumar, Machine Learning from Theory to Algorithms: An OverviewJournal of Physics: Conference Series: Volume
1142 , Dec 2018.
[2] A report by Royal Society, April 2017, “Machine learning: the power and promise of computers that learn by example “, Page 41 ISBN: 978-1-78252-259-
1.and Information Technologies, Vol. 7 (3) , 2016, 1174- 1179.
[3] PariwatOngsulee 15th International Conference on ICT and Knowledge Engineering (ICT&KE), nov 2017,DOI: 10.1109/ICTKE.2017.8259629.
[4] Shalev-Shwartz S, Ben-David S. 2014 Understanding machine learning: from theory to algorithms. Cambridge, UK: Cambridge University Press.
[5] Markoff J. 2015 A learning advance in artificial intelligence rivals human abilities. New York Times. 10 December 2015. See
https://www.nytimes.com/2015/12/11/science/an-advance-in-artificial-intelligence-rivals-human-vision-abilities.html.
[6] Sandhya N. dhage, Charanjeet Kaur Raina, “A review on Machine Learning Techniques”, March 16 Volume 4 Issue 3 , International Journal on Recent and
Innovation Trends in Computing and Communication (IJRITCC), ISSN: 2321-8169, PP: 395 – 399.
[7] IBM. IBM Watson Health. See http://www.ibm.com/watson/health/oncology.
[8] A report by Royal Society, April 2017, “Machine learning: the power and promise of computers that learn by example “, Page 41 ISBN: 978-1-78252-259-
1.and Information Technologies, Vol. 7 (3) , 2016, 1174- 1179.
[9] Condlifee, J. 2016 An Amazon drone has delivered its first products to a paying customer. MIT Technology Review.
See https://www.technologyreview.com/s/603141/an-amazon-drone-has-delivered-its-first-products-to-a-payingcustomer
[10] Joe Lemley; ShababBazrafkan; Peter Corcoran, Deep Learning for Consumer Devices and Services: Pushing the limits for machine learning, artificial
intelligence, and computer vision.IEEE Consumer Electronics Magazine ( Volume: 6, Issue: 2, April 2017),Page(s): 48 – 56.
[11] Shekhawat V.S., Tiwari M., Patel M. (2021) A Secured Steganography Algorithm for Hiding an Image and Data in an Image Using LSB Technique. In:
Singh V., Asari V.K., Kumar S., Patel R.B. (eds) Computational Methods and Data Engineering. Advances in Intelligent Systems and Computing, vol 1257.
Springer, Singapore. https://doi.org/10.1007/978-981-15-7907-3_35.
[12] H. Gupta and M. Patel, "Study of Extractive Text Summarizer Using The Elmo Embedding," 2020 Fourth International Conference on I-SMAC (IoT in
Social, Mobile, Analytics and Cloud) (I-SMAC), Palladam, India, 2020, pp. 829-834, doi: 10.1109/I-SMAC49090.2020.9243610.
[13] Shailja Joshi and Mayank Patel. “Natural Language Processing for Classifying Text Using Naïve Bayes Model.”, PAIDEUMA JOURNAL, Volume 13,
Issue 10, ISSN NO : 0090-5674, Page No: 72-77, DOI : 11.3991/Pjr.V13I10.85307
[14] NeelamBadi, Mayank Patel and Amit Sinhal (2019). The Role of Fuzzy Logic in Improving Accuracy of Phishing Detection System. International Journal
of Innovative Technology and Exploring Engineering, Volume-8 Issue-8, ISSN: 2278-3075, pp.3162-3164.
[15] Menaria H.K., Nagar P., Patel M. (2020) Tweet Sentiment Classification by Semantic and Frequency Base Features Using Hybrid Classifier. In: Luhach A.,
Kosa J., Poonia R., Gao XZ., Singh D. (eds) First International Conference on Sustainable Technologies for Computational Intelligence. Advances in
Intelligent Systems and Computing, vol 1045. Springer, Singapore. https://doi.org/10.1007/978-981-15-0029-9_9
[16] [16] K. C. Giri, M. Patel, A. Sinhal and D. Gautam, “A Novel Paradigm of Melanoma Diagnosis Using Machine Learning and Information Theory," 2019
International Conference on Advances in Computing and Communication Engineering (ICACCE), Sathyamangalam, Tamil Nadu, India, 2019, pp. 1-7, doi:
10.1109/ICACCE46606.2019.9079975.
[17] PReDicT (Predicting Response to Depression Treatment). See http://predictproject.p1vitalproducts. com/
[18] Condon, C. 2016 Quest for robo-Yellen advances as computers gain on rate setters. Bloomberg. See https://www. bloomberg.com/news/articles/2016-05-
24/quest-for-robo-yellen-advances-as-computers-gain-on-rate-setters
[19] DeepMind. 2016 Press release: DeepMind AI reduces Google data centre cooling bill by 40%. See https://deepmind. com/blog/deepmind-ai-reduces-
google-data-centre-cooling-bill-40/

©IJRASET: All Rights are Reserved 1310