Advances in Applied Science Research

ISSN: 0976-8610

Open Access Review Article

Face Recognition and Machine Learning

Sameer Sadiq Sofi*, Chandikaditya Kumawat, Rafi Ahmad Khan
Department of Mechanical Engineering, University of Kashmir, Srinagar, Jammu andKashmir, India

Abstract
Face recognition has become an important area of research from the last decade as there is much
need of security for data as well as physical assets of almost every organization. Today is the age of
data and with increasing voluminous amounts of data day by day, there is need of some sophisticated
system to manage this vast data. Moreover, these organizations need some robust system to secure
as well as manage these data resources so that only authorized users can access them. Face
recognition system can be used on a large scale to meet this demand especially by using some
techniques of machine learning. ML techniques promise of producing accurate estimates. In this
paper, we are explaining various machine learning techniques and also explain working of machine
learning model in context of face recognition.

Keywords: FRS; Machine learning; Supervised; Un supervised; AI; Expert system

INTRODUCTION LITERATURE REVIEW

Many organizations and industrial sectors have been using In this section, we will discuss ML and its theories and
machine learning from the last two decades and new smart continue with more popular and useful neural network
applications are being developed using this process and is on techniques.
rise. Today machine learning is affecting almost every field
such as education, finance, health, banking, military and many Machine Learning
more. Machine Learning (ML) is used in the context to learn
Learning is the process of modifying and improving skill or a
from the data. It is being utilized by enabling machines to
program based on past experience, automatically without any
handle data more efficiently and in a proper way. When
external human assistance. It is cognitive process through
humans cannot quickly interpret or extract information after
which knowledge or a skill is acquired. Artificial Intelligence
viewing data, then ML is applied [1]. ML is used to design and
(AI) played a great role in evolving machine learning by
develop programs that learn and improve continuously as
applying its computational learning theory while studying
applied to new data. As the datasets are increasing in size, ML
pattern recognition in images. Machine learning is a subset of
is also getting its scope wider. Research is on to enable
artificial intelligence that helps machines to perform very
machines to learn themselves [2,3].
crucial tasks and important applications by making data
driven decisions without need of explicitly programming
them. In this procedure, a model is constructed from example

Received: 25-May-2022 Manuscript No: AASRFC-22-13652

Editor assigned: 27-May-2022 PreQC No: AASRFC-22-13652
Reviewed: 10-June-2022 QC No: AASRFC-22-13652
Revised: 10-October-2022 Manuscript No: AASRFC-22-13652
Published: 17-October-2022 DOI: 10.36648/0976-8610.13.11.98

Corresponding author Sameer Sadiq Sofi, Department of Mechanical Engineering, University of Kashmir, Srinagar, Jammu
and Kashmir, India; E-mail: sofi.sam4@gmail.com
Citation Sofi SS, Kumawat C, Khan RA (2022) Face Recognition and Machine Learning. Adv Appl Sci Res. 13:98.
Copyright © 2022 Sofi SS, et al. This is an open-access article distributed under the terms of the Creative Commons
Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original
author and source are credited.

© Under License of Creative Commons Attribution 4.0 License

Volume 13 • Issue 11 • 98
This article is available in: https://www.primescholars.com/advances-in-applied-science-research/archive.html
Page 2
Sofi SS, et al.

inputs to perform data driven choices or predictions without Categories of Machine Learning
following static program instructions. The model improves
over time as it experiences on exposing to new data. With the Machine learning can be broadly categorized into four sub
help of Machine Learning deficiencies of manual knowledge classes: These sub classes have their own respective
acquisition techniques can be overcome by automating the objectives and techniques which make them capable to
learning process. We can broadly classify machine learning implement different types of learning and mechanisms.
into two groups or methods: a) inductive and b) deductive.
• Supervised learning.
Deductive learning use existing knowledge and facts. It then
creates new knowledge from old facts. Inductive learning • Un-supervised learning.
generalizes and extracts new rules and patterns from big • Semi-supervised learning.
datasets rather than operating the existing knowledge. • Reinforcement learning.
Supervised Learning
Evolution of ML
It is that type of Machine Learning in which after giving new
Since machine learning is a sub filed of artificial intelligence, data, a pre defined set of ‘training examples’ enable its ability
its history dates back to the 1950's which were cognitive to reach an accurate conclusion. In Supervised Learning,
science days. In that era, birth of many general techniques ‘labels’ are used for data and given as examples. After that a
took place and at that time bioscience was more popular and suitable learning algorithm is applied to these ‘example label’
development of perceptron was a landmark, which was pairs one by one and after predicting the label for each
modeled after neurons [4]. Later on work on neural networks example by the algorithm, feedback is received and thus it
began. The period of 1970’s was featured as more practical depicts whether the right choice is predicted or not. With
programs and algorithms were developed and this time, the algorithm will learn to approximate the exact nature
achievement of designing algorithms was characterized with of the relationship between labels and examples. After some
usage of symbolic techniques [5]. Many useful discoveries time, when the algorithm attains full training, it then becomes
belonging to that period included ‘learning of knowledge capable to identify and observe a new example and predicts a
acquisition bottleneck’, Buchanan and Mitchell's Meta- better label for it. In supervised machine learning algorithms,
Dendral, and results of diagnosing soyabean. In 1980’s, there is need of external assistance and the input dataset is
‘version spaces’ was developed after increased analysis and divided into training and testing dataset. The training dataset
evaluation of various Learning techniques [6], and much needs to be predicted as the algorithms of this kind have
achievement was creation of decision tree algorithm. variable output and then dataset is classified. In the same
Probably Approximately Correct (PAC) learning was defined by way, mostly all algorithms learn some kind of patterns from
Valiant in 1984 and the back propagation algorithm removed the training dataset and then they put on them to the test
many limitations of perceptron [7]. Different learning dataset for classification or prediction [8]. The classification
techniques such as explanation based learning, case based problem is one of the standard making of supervised learning,
learning and speedup learning were deeply analysed and here the learner is requisite to learn, in other words, to
became popular. Significant progress was on rise in decision approximate the behaviour of a function that connects a
tree and rule learning at the end of this period. During 1990’s vector into one of several classes by looking at several input/
organizational data was made available and compared with output examples of the function [9]. Here knowledge is
algorithms. Meanwhile data mining emerged and this made it induced from already known observations.
possible to compare various statistical records. Multi
relational learning and Kernal methods were designed and Decision tree: It is a predictive modelling technique in the
analysed. Reinforced learning, bayesian learning, automatic field of machine learning and statistics which can be used to
bias selection, adaptive software agents, voting, bagging, model the underlying data patterns. It is also an example of a
boosting, inductive logic programming and applications to classification algorithm. Classification algorithms are utilized
robotics became popular. in solving problems ranging from credit card theft detection to
diagnosing patients with heart defects by recognizing
2000 and onwards: This period showed interaction between distinctive patterns in a dataset and classifying activity based
computational learning theory, symbolic machine learning on that retrieved information [10].
statistics, pattern recognition and neural networks. Various
new applications for ML techniques such as knowledge Support Vector Machines (SVMs): It works on the principle of
discovery in databases; robot control; language processing ‘margin calculation’. Basically margins are drawn between the
and combinatorial optimization were on rise. Also support classes via some non-linear mapping ‘input vectors’ into a
vector machines, statistical relational learning and ensembles high dimensional feature space. The maximum separation
were touching new heights. Learning ensembles, learning between decision classes is given by maximum margin hyper
complex stochastic models, scaling up Supervised Learning plane and in this way classification errors are minimized. The
Algorithms (SLA) were developed to improve accuracy. closest training examples to the maximum margin hyper plane
are termed as the ‘support vectors’.

Volume 13 • Issue 11 • 98
Page 3
Sofi SS, et al.

Neural networks: An artificial neural network can be theory has led to development of neural models that are
constructed for a specific application, such as data used for pattern recognition and unsupervised learning.
classification or pattern recognition, through a learning Learning of stable recognition categories have been made
process. These networks can be utilized with their ability to possible by these models. A variety of cognitive and brain
derive meaning from complicated or imprecise data and thus data models are being explained with the help of ART
are able to extract patterns as well as detect trends that systems.
otherwise are too complex to be noticed by either humans or
Clustering: In clustering, unlike classification the groups are
other computer methods. Hence, ANNs can be used as
not predefined. Here, the grouping is performed by finding
classifiers for various security applications and other crucial
similarities between data. A similarity metric is defined for the
tasks.
concerned data. A nested set of clusters is created in
Rough set: Rough set theory has been employed in hierarchical clustering where the number of clusters is not
discovering knowledge in experimental databases and known beforehand.
datasets [11,12]. In this technique an upper and a lower
approximation of a set are carried out. According to Semi-supervised
Obersteiner and Wilk, the ‘rough sets theory’ is measured on
the assumption that some information is associated with Semi supervised learning: Here strength of both
every object of the considered universe (data, knowledge), supervised and unsupervised learning is combined. When
expressed in the form of some attributes that is used for unlabeled data is already present, effective results are
description of the object under consideration [13]. produced and getting the labeled data is a difficult
process. There are various categories of semi supervised
Un-supervised Learning learning algorithms which combine both labelled and
Here labels are not used, instead the learning algorithm unlabeled examples to generate an appropriate function or
receives a lot of data and after using the tools it understands classifier.
the properties of the data. From that point the algorithm is
able to group, cluster, and organize the data in such a way Reinforcement Learning
that a human or some other intelligent algorithm can come In this type of learning decisions are made based on what to
and make sense of the newly organized data. In this way the do, how to map situations to actions so as to maximize a
model becomes able to learn through observations, it then numerical reward signal. Prior knowledge is not available to
finds structures in the data. When the model is given a learner about which actions to take. Reinforcement learning is
dataset, clusters are created by automatically finding the applicable to limited to a small set of real world situations,
relationships and patterns in the dataset. Unsupervised until it’s been given a situation. It is the responsibility of the
learning can mostly be used for clustering and feature learner who needs to discover which actions yield the most
reduction and as it is data driven and no usage of labels is reward by trying them. Reinforcement learning has two most
performed, the outcomes are controlled by the data and the important distinguishing features ‘trial and error search’ and
way it is formatted. ‘delayed reward’ and these are the its two most.
Principal component analysis: PCA is used to reduce Q-learning: It is a type of reinforcement learning that does not
dimensions between the data to make the computations solely depend on model. Q-learning uses the information
faster and easier. Working of PCA can be understood by taking observed to approximate the optimal function from where the
example of 2D data. When we plot data in a graph, it will take optimal policy can be constructed. In the initial stages the
up two axes. When we apply PCA on that data; the data then learning rate is high and it gradually decreases so the learner
will be 1D. needs to perform each state action infinitely. The Q-learning
Self-Organizing Maps (SOM): SOM network is an has requirements that the sum of the learning rates goes to
unsupervised network and is mainly used as a clustering infinity and that the sum of the squares of the learning rates is
technique when no training data are available. The principle finite.
of clustering algorithm is that the degree of data pattern
similarity within a cluster is maximized, while the similarity How does Machine Learning Work?
which these patterns have with the patterns belonging to A training data set is used to create a model to train machine
different/other clusters is minimized. SOM technique has learning algorithm. This model is then used by ML algorithm
recently been used for the visualization and analysis of to make a prediction as it encounters new data. After that the
symbolic and text data [14]. model is gradually tested and evaluated for accuracy and if
Adaptive Resonance Theory (ART): Stephen Grossberg and there acceptable accuracy then the ML algorithm is deployed
Gail Carpenter introduced the Adaptive Resonance Theory otherwise the model is trained further to attain accuracy with
(ART) for human cognitive information processing [15]. This an augmented training data set again and again (Figures 1 and
2).

Volume 13 • Issue 11 • 98
Page 4
Sofi SS, et al.

A machine learning system has three major building blocks:

The model, the parameters, and the learner
• Model is used for making predictions.
• The parameters are the factors that the model considers
to make predictions.
• The learner’s job is to make adjustments in the parameters
and the model to align the predictions with the actual
results.

Figure 1: Machine Learning Process (MLP). Here we will explain each step of a machine learning system
through a ‘known’ and unknown face example to understand
Elaborating Each Step how machine learning works. A machine learning model
predicts whether a face is known or unknown. The
parameters selected are the facial features like presence of
eyes, nose, cheekbones, lips etc. of the human face. The first
step is:

Gathering Data
The quantity and quality of the data dictate how accurate the
model is. The outcome of this step is generally a
representation of what will be used for training. We can also
use pre collected data in this step. In this case, the data we
collect will be feature vectors containing human face
parameters.
Figure 2: Elaborating each step of machine learning.

Table 1: Indicating feature vector and match percentage versus classification of face.
Face encoding (feature vector) Match (%) known or unknown

170 30-55 Unknown

212 60-90 Known

220 65-95 Known

This will yield a table of ‘face encoding’, ‘match %’, and determination of what a face is, independent of what face
whether it is known or unknown. This will be our training came before or after it. Here we do any pertinent visualization
data. of our data, to help to see if there are any relevant
relationships between different variables you can take
Preparing Data advantage of, as well as show you if there are any data
imbalances. For example, if we collected way more data
• Spat/wrangle input data and further prepare it for points about known than unknown, the model we train will
training. be biased toward guessing that virtually everything that it
• Wrangled data needs to be clean with removing sees is known, since it would be right most of the time.
duplicates, correct errors, deal with missing values, However, in the real world, the model may see known and
normalization, and data type conversions, etc. unknown an equal match, which would mean that guessing
• After that we need to randomize data to erase the effects “known” would be wrong half the time. Also we split the data
of the particular order in which we collected or otherwise in two parts, the first part, used in training our model, will be
prepared our data. the majority of the dataset. The 2nd part will be used for
• Then data is visualized to help to detect relevant evaluating our trained model’s performance. Same data that
relationships between variables or class imbalances, or in the model was trained on for evaluation cannot be used, since
other words perform other exploratory analysis. it could then just memorize the “questions”. Sometimes the
• Lastly data is split into training and evaluation sets. collected data needs other forms of adjusting and
manipulation. Processes like de duping, normalization, error
In this step, we load our data into a suitable place and correction, etc., these would all happen at the data
prepare it for use in our machine learning training. The data is preparation step.
first put all together, and then the order of data is
randomized. In our aim to classify face, we make a

Volume 13 • Issue 11 • 98
Page 5
Sofi SS, et al.

Choosing a Model
As different algorithms exist and are for different tasks, we
need to choose the right one. Researchers and data scientists
have created different models over the years and some are
very well suited for image data, others for sequences like text,
or music, some for numerical data, and others for text based
data. In our case, since we only have 2 features, known/
unknown and match %, we can use a small linear model,
which is a fairly simple to do our job.

Train the Model

• To train the model, is to answer a question or make a
prediction correctly as often as possible.
• Considering the linear regression example: Algorithm here
needs to learn values for m (or W) and b (x is input, y is
output).
• So each iteration of the process is a training step.
In this step, we will use our data to incrementally improve our
model’s ability to predict whether a given face is known or
unknown. This is performed on a much smaller scale with our
faces images. In particular, the formula for a straight line is’
y=m × x+b’, where x is the input, m is the slope of that line, b is
the y-intercept, and y is the value of the line at the position x. Figure 3: The values in W and b such that, we will have more
These values that we have available to us for adjusting, or correct predictions.
‘training’, are m and b. There is no other way to affect the
position of the line as the only other variables are x, our This process is then repeated and each iteration or cycle of
input, and y, our output. In machine learning, there can be updating the weights and biases is termed as one ‘training
many m’s since there may be many features. The collection of step’. In case of our dataset when we first start the training,
these m values is usually formed into a matrix, that we will it’s like we drew a random line through the data. Then with
denote W, for the “weights” matrix. Likewise for b, we arrange each step of the training progresses, the line moves step by
them together and call that the biases. In training process step closer to an ideal separation of the known and unknown.
initializing some random values for W and b are involved and
attempt is made to predict the output with those values. As it Error Measurement
can be imagined, the model does pretty poorly. But by After training the model on a defined training set, it needs to
comparing our model’s predictions with the output that it be checked for any discrepancies or errors. To do this task, we
should produce and adjusting the values in W and b such that, use a fresh set of data to accomplish it. The result of this test
we will have more correct predictions (Figure 3). would be one of these four outcomes
• True positive: This outcome occurs when the model
predicts the condition when it is present.
• True negative: When the model does not predict a
condition when it is absent, this result occurs.
• False positive: The model predicts ‘such a condition when
it is absent’.
• False negative: The outcome is False Negative when the
model does not predict a condition when it is present.
Keeping the above outcomes in view, the sum of FP and FN is
the total error in the model.

Volume 13 • Issue 11 • 98
Page 6
Sofi SS, et al.

Noise Management data, but might have more significant error while processing
new data; we call this as ‘under fitting’. But if the hypothesis
• Further, the hypothesis then created will have a lot more is too complicated to accommodate the best fit to the training
errors because of the presence of the noise. Noise is result, it might not generalize well; this is the case of ‘over
defined as unwanted anomalies that disguise the fitting’. In either case, the results are fed back to train the
underlying relationship in the data set and weakens the model further.
learning process. This noise may occur due to a lot of
below reasons:
Parameter Tuning
• If dataset for training is large.
• When errors are already in input data. • Parameter tuning refers to hyper parameter tuning, which
• When data labelling has errors. tune model parameters for improved performance.
• Unobservable attributes that are not considered in the • Simple model hyper parameters may include following sub
training set due to lack of data and might affect the processes: What is the number of training steps, learning
classification process. rate, initialization values and distribution?

To approach a machine learning problem here that is the After evaluation is done, we might want to see if we can
known/unknown and match percentage for instance and for further improve our training in any way. This can be done by
the sake of simplicity, we have considered only two tuning our parameters. There were a few parameters we
parameters. But in reality, we may have to consider hundreds implicitly assumed when we did our training, and now we can
of parameters and a broad set of learning data to solve a go back and test those assumptions and try other values. To
machine learning problem. To keep the hypothesis as simple understand this, we can explain as how many times we run
as possible, we can accept a certain degree of training error through the training dataset during training. This means that
due to noise. we can “show” the model our full dataset multiple times,
rather than just once. This can sometimes lead to higher
accuracies (Figure 4).
Evaluate the Model
• We use some metric or combination of metrics to
measure objective performance of the model.
• The model is tested against previously unseen data.
• In the real world, this unseen data is meant to be
somewhat representative of model performance but it still
helps tune the model as opposed to test data, which does
not.
• Good train/eval split? 80/20, 70/30, or similar, this is
depending on domain, data availability, dataset
particulars, etc.
As our training is complete, now we should see how the
model is any good, this is done using evaluation. This is where
that dataset that we set aside earlier comes into play.
Evaluation allows us to test our model against data that has
never been used for training and this metric enables us to see,
Figure 4: An experimental process of Face Recognition and
‘how the model might perform against data that it has not yet
Machine Learning.
seen’. We do this to make the model a representative of how
it might perform in the real world. A good rule of thumb I’ can Another parameter is “learning rate”. This refers to how far
be used for a ‘training evaluation split’ somewhere on the we shift the line during each step, based on the information
order of 80/20 or 70/30 of which much depends on the size of from the previous training step. These values all play a role in
the original source dataset. Perhaps you don’t need as big of a how accurate our model can become, and how long the
fraction for the evaluation dataset, if there is a lot of data. training takes. For more complex models, initial conditions
can play a significant role in determining the outcome of
Testing and Generalization training. Differences can be seen depending on whether a
model starts off training with values initialized to zeroes
An algorithm or hypothesis might be possible to fit well to a versus some distribution of values, which leads to the
training set and might fail when applied to another set of data question of which distribution to use. We typically refer these
outside of the training set, and then only data is the way to parameters as “hyper parameters”. The raining or adjustment
judge this, i.e., testing. Further, generalisation refers to how of these hyper parameters remains a bit of an art, and is more
well the model predicts outcomes for a new set of data; of an experimental process that heavily depends on the
therefore, it is essential to figure out if the algorithm is fit for specifics of your dataset, model, and training process.
new data. As we fit a hypothesis algorithm for maximum
possible simplicity, it might have less error for the training

Volume 13 • Issue 11 • 98
Page 7
Sofi SS, et al.

Make Predictions Gaming sector has increasing bright future in learning through
Reinforcement Learning.
The class labels that were withheld from the model till this
point can further be used to test the model as test set, a Industrial simulation: Many robotic applications like ‘think
better a better approximation of how the model will perform assembly lines’ for industrial simulation can learn to complete
in the real world can be traced. Machine learning is using data their tasks without having to hardcode their processes. With
to answer these questions. So Prediction, or inference, can be IS, severe failures in industries can be saved and tasks can be
termed as the step where we get to answer some questions. cost effective.
At this point of time and all this work, the value of machine Resource management: Reinforcement learning can be used
learning is realized. In this approach we can finally use our to perform navigation of complex environment and it can
model to predict whether a given face is known or unknown, handle the necessity to balance certain requirements. For
given its feature vector and match percentage. instance, google’s data centers applied reinforcement learning
to balance the need to satisfy our power requirements and
DISCUSSION also efficiently and effectively cutting major costs. In bio
surveillance, machine learning is applied to detect and track
major disease outbreaks. For example, the RODS project
Machine Learning Applications
involves real time collection of admissions reports to
Machine learning has achieved much progress in tackling real emergency rooms across western Pennsylvania, and the use
world applications. The following achievements present the of machine learning software to learn the profile of typical
utilization and applications of different types of machine admissions so that it can detect anomalous patterns of
learning’s: symptoms and their geographical distribution over cost
effective lines and intend of major accuracy.
Advertisement popularity: Advertisements are made more
effective with the help of supervised learning by providing
clicking ability more attractive. Also with a learning algorithm, CONCLUSION
match between advertisement and placement on a website is
made more effective. This study discussed various techniques of machine learning
that have played a major role in improving the process of face
Spam classification: A modern email spam filter is managed recognition. But still we face various challenges in this field
with the help of supervised algorithm. With the usage of and it needs much more to be done as the loopholes are
labels indicating spam, not spam, this system learns how to evident in many aspects. There is a need to improve upon
pre emotively filter out malicious emails so the user is saved those challenges and come up with a more robust system of
from dangers and other hacking staff. face recognition and this is possible through proper and
Recommender systems: In youtube or netflix, different video efficient use of intelligent machine learning techniques
recommendation systems are often kept in the unsupervised especially neural networks. Now research is in progress and
domain. Different parameters of videos such as their new techniques are being devised for extracting rules from
duration, their genre etc, watch history of different users, neural networks, and combining neural networks with other
watching similar videos and their relationship in the data intelligent techniques like genetic algorithms, fuzzy logic and
prompt us such a suggestion. expert systems to arrive at appropriate conclusions and
solutions. In most cases, neural networks perform well than
Buying habits: Buying and selling habits of online users are the traditional classification techniques with which they are
contained in a database somewhere and that data is being compared. Deep learning, due to its vast features and high
bought and sold actively at this time. An unsupervised learning capacity, has tremendous scope in future and it can
algorithm is applied to group customers into similar help humans in achieving real success in securing their
purchasing segments. This way companies market to these valuable assets and overcome large scale destructions caused
grouped segments and can even resemble recommender by natural calamities by getting valuable information in
systems. advance. Such technology can help in making day to day life
Grouping user logs: Unsupervised learning can be used to comfortable and secured in this fast and smart world.
group user logs and issues and with this learning feature
companies can design and use identifying central themes to
issues their customers face and rectify issues, through
REFERENCES
improving a product or designing an FAQ to handle common 1. Richert W, Coelho LP (2013) Building Machine Learning
issues. An issue with a product or submitting a bug report Systems with Python. Packt Publishing, Birmingham, UK.
often fed to an unsupervised learning algorithm to cluster it 1-290.
with other similar issues. 2. Welling M (2010) A First Encounter with Machine
Video games: Various online games are mostly played by Learning. 1-93.
using reinforcement learning. Alpha zero and alpha go are 3. M Bowles (2019) Machine Learning in Python: Essential
google’s reinforcement learning application which are learned Techniques for Predictive Analytics. 2nd Ed. Wiley
to play the game go. Another example is ‘our mario’ game. Publishers, New York, United States. 1-368.

Volume 13 • Issue 11 • 98
Page 8
Sofi SS, et al.

4. Rosenblatt F (1957) The Perceptron: A Probabilistic 11. Grossberg S (1976) Adaptive pattern classification and
Model for Information Storage and Organization in the universal recoding, I: Parallel development and coding of
Brain. Psychol Rev. 65(6):386-408. neural feature detectors and II: Feedback, expectation,
5. Mitchell TR (1982) Motivation: New directions for theory olfaction, and illusions. Biol Cybern. 23(4):187-202.
and research. Acad Manage Rev. 7(1):80-88. 12. Zhu GX (2009) Introduction to Semi Supervised Learning.
6. Kotsiantis S (2007) Supervised Machine Learning: A Synthesis Lectures on Artificial Intelligence and Machine
Review of Classification Techniques. Informatica. Learning. Morgan and clay pool publishers. Madison,
249-268. USA. 1-130.
7. Chan PK, Stolfo SJ (1999) Toward Scalable Learning with 13. Sutton SR, Barto AG (1998) Reinforcement Learning: An
Non-uniform Class and Cost Distributed data mining in Introduction. Massachusetts: MIT Press, London,
credit card fraud detection. IEEE Intell Syst.14(6):67-74. England. 1-352.
8. Pawlak Z (1985) Rough sets. Int J Comp Sci. 11(1): 14. Sutton RS (1992) Introduction: The Challenge of
341-356. Reinforcement Learning. Mach Learn. 173(1):225-227.
9. Skowron A (1990) The rough sets theory and evidence 15. Mansour ED, Yishay E (2003) Learning Rates for Q-
theory. Fundam Inform. 13(3):245-262. learning. Mach Learn Res. 5:1-25.
10. Obersteiner M, Wilk S(1999) Determinants of Long term
Economic Development: An Empirical Cross country
Study Involving Rough Sets Theory and Rule Induction.
Vienna: Insititute for Advanced Studies.

(MRPFT) Volume 13 • Issue 11 • 98