BIO Web of Conferences 97, 00133 (2024) https://doi.org/10.

1051/bioconf/20249700133
ISCKU 2024

An overview of machine learning

classification techniques
Amer F.A.H. ALNUAIMI1* and Tasnim H.K. ALBALDAWI1
1Dept. of Mathematics, College of Science, University of Baghdad, Baghdad, 10071, Iraq

Abstract. Machine learning (ML) is a key component within the

broader field of artificial intelligence (AI) that employs statistical
methods to empower computers with the ability to learn and make
decisions autonomously, without the need for explicit
programming. It is founded on the concept that computers can
acquire knowledge from data, identify patterns, and draw
conclusions with minimal human intervention. The main categories
of ML include supervised learning, unsupervised learning, semi-
supervised learning, and reinforcement learning. Supervised
learning involves training models using labelled datasets and
comprises two primary forms: classification and regression.
Regression is used for continuous output, while classification is
employed for categorical output. The objective of supervised
learning is to optimize models that can predict class labels based on
input features. Classification is a technique used to predict similar
information based on the values of a categorical target or class
variable. It is a valuable method for analyzing various types of
statistical data. These algorithms have diverse applications,
including image classification, predictive modeling, and data
mining. This study aims to provide a quick reference guide to the
most widely used basic classification methods in machine learning,
with advantages and disadvantages. Of course, a single article
cannot be a complete review of all supervised machine learning
classification algorithms. It serves as a valuable resource for both
academics and researchers, providing a guide for all newcomers to
the field, thereby enriching their comprehension of classification
methodologies.

1 INTRODUCTION
Machine learning is an interdisciplinary field that draws its foundations from
statistical theory, computer science, and other related domains. Its meteoric rise in
* Corresponding author: [email protected]

© The Authors, published by EDP Sciences. This is an open access article distributed under the terms of the Creative Commons Attribution
License 4.0 (https://creativecommons.org/licenses/by/4.0/).
BIO Web of Conferences 97, 00133 (2024) https://doi.org/10.1051/bioconf/20249700133
ISCKU 2024

popularity in recent years has positioned it as a leading force in technological

advancements. Previously referred to by different terms such as statistical learning
and pattern recognition, machine learning now encompasses a wide array of
meticulously studied, refined, and honed methods [1]. The close relationship between
statistics and machine learning is evident, with statistics providing the mathematical
underpinning for creating interpretable statistical models that unveil concealed
insights within intricate datasets.
At the core of machine learning lies the concept of self-learning, a process that
employs statistical modeling, a subfield of mathematics dealing with the discovery of
relationships between variables to predict outcomes [1]. This self-learning capability,
achieved through statistical modeling and data-driven insights, has become a pivotal
feature of machine learning [2]. The credit for coining and defining machine learning,
as it stands today, goes to Arthur Samuel [2]. Even though his initial definition did
not explicitly mention self-learning, this concept has evolved to become a central
tenet of machine learning [3], [4]. Self-learning involves using statistical modeling to
discern patterns and enhance performance based on data and empirical knowledge,
all without the need for explicit programming commands. This approach has been
widely embraced for nearly six decades.
Computer science serves as the bedrock of machine learning, data mining, and
computer programming, encapsulating the essence of these disciplines. It finds its
niche within the broader realm of data science, which is dedicated to extracting
valuable insights from vast data volumes through computer-driven methodologies.
Expanding further, artificial intelligence (AI) breathes life into machine learning and
data science, encompassing the development of intelligent machines that emulate
cognitive abilities [4].
The convergence of machine learning, statistical learning theory, and data science
resides in their shared quest for data processing, the construction of adaptive models,
and precise predictions. The term "data science" underscores the development of
robust machine learning and computational techniques that can tackle the challenges
posed by large-scale data analysis [5]. As machine learning continues its
extraordinary ascent, it has reshaped career trajectories, propelling the "Data
Scientist" role to the forefront as the most sought-after profession of the 21st century.
While machine learning finds its applications across various domains,
classification in supervised learning stands out as a pivotal discipline where its
capabilities truly shine. It enables the precise prediction and empowers informed
decision-making. Understanding classification techniques is fundamental for
achieving dependable results in diverse domains. Effective data management and a
deep comprehension of classification algorithms facilitate valuable insights
extraction and knowledge discovery [3]. These techniques find wide-ranging
applications, including data categorization, prediction, and pattern recognition [6].
Machine learning, with its capacity to learn from data and make predictions, is
reshaping various industries and sectors. It equips us with the ability to make sense
of intricate datasets, identify patterns, and make informed decisions. By grasping the
definitions, advantages, limitations, and applications of classification techniques in
machine learning, we can unlock the full potential of this technology and explore new
horizons in research and development [7], [8].

2 MACHINE LEARNING

2
BIO Web of Conferences 97, 00133 (2024) https://doi.org/10.1051/bioconf/20249700133
ISCKU 2024

Machine learning (ML) is a branch of computer science and AI that employs

statistical methods to empowers computers to acquire knowledge and improve their
performance without requiring direct programming. It is founded on the concept that
computers can acquire knowledge from data, identify patterns, and draw conclusions
with minimal human intervention [1], [4], [9].

2.1 Types of Machine Learning

Machine learning algorithms are trained by utilizing labelled, unlabelled or hybrid
datasets. Consequently, the nature of the requisite data for a particular task dictates
the archetype of machine learning model that is formulated. Consequently, four
fundamental categories of machine learning have been established developed as [3],
[5], [9], [10]:

2.1.1 Supervised learning

Supervised learning (SL) is a machine learning approach that leverages labelled
data to educate a system in forecasting outcomes based on its training. It closely
mimics the process of human learning under the guidance of an instructor, employing
specific instances to deduce overarching principles. SL is typically divided into two
main categories [2], [5], [9]:
Regression: Regression is a term used in statistics, which is a type of statistical
analysis that aims to understand the relationship between a dependent variable
(response variable) and one or more independent variables (predictors) such as in
market trends or weather forecasting. The most common type is linear regression [5],
[7].
Classification: Classification is a SL technique that involves categorizing data
into distinct classes. It is a recursive process that recognizes and groups data objects
into pre-defined categories or labels [11]. This technique is used to predict the
outcome of a given problem based on input features. It can be applied to structured
or unstructured data, and the classes are commonly known as target, label, or
categories. The aim of classification is to assign an unknown pattern to a known class.
For example, classifying emails as "spam" or "not spam" is a common application of
classification [12].
Both the Classification and Regression algorithms can be used for forecasting in
machine learning and operate with the labelled datasets. But the distinction between
classification vs regression is how they are used on particular machine learning
problems.

2.1.2 Unsupervised learning

Unsupervised learning involves training the machine without knowing the output,
using only input samples or labels. It discovers patterns in data and creates its own
data clusters. There are two main types and most common of unsupervised learning
algorithms, clustering and association analysis. This technique is useful for
identifying unknown patterns in data, such as such as recommendation engines on
online stores that rely on unsupervised machine learning, specifically a technique
called clustering [9], [13].

3
BIO Web of Conferences 97, 00133 (2024) https://doi.org/10.1051/bioconf/20249700133
ISCKU 2024

2.1.3 Semi-Supervised Learning

Semi-Supervised Learning (SSL) is a powerful technique that combines
supervised and unsupervised learning to improve learning accuracy. It addresses the
challenge of limited labelled data by leveraging the abundance of unlabelled data [9],
[13]. SSL finds applications in diverse fields like speech analysis, web content
classification, protein sequence classification, and text document classifiers [14].

2.1.4 Reinforcement Learning

Reinforcement Learning (RL) is a unique form of machine learning that enables
machines and software agents to determine optimal behaviour in order to achieve
desired outcomes [14]. Unlike traditional approaches, RL relies on trial-and-error
experiments by receiving feedback that reinforces favourable outputs and discourages
unfavourable ones. As the agent accumulates experience, it improves upon itself and
becomes better at making decisions to achieve its objective [15], where an agent
interacts with a dynamic environment, receiving rewards and punishments, to achieve
a specific goal without explicit guidance from a teacher [13].
Although sometimes considered a subset of semi-supervised learning, RL is
widely recognized as a distinct form of machine learning studied across various
disciplines due to its generality [10]. These disciplines include game theory, control
theory, operations research, information theory, simulation-based optimization,
multi-agent systems, swarm intelligence, statistics, and genetic algorithms [7], [9].
One practical example of RL is autonomous driving.

3 TYPES OF CLASSIFICATION
The field of classification encompasses four primary task types. These include
binary classification, multi-class classification, multi-label classification, and
imbalanced classification. Resources and references related to these categories can
be found in [10], [12], [16].

4 SUPERVISED LEARNING CLASSIFICATION

ALGORITHMS
SL algorithms are used in AI to classify data into different classes or groups [17].
They are trained based on data or observations, and new data is classified based on
its class. Typically, a dataset is divided into training and testing sets for accuracy
evaluation [18]. Classification predictive modelling aims to map input variables to
discrete output variables using function f. Classification algorithms transform input
data into desired output using mathematical and logical models. There are numerous
classification algorithms and techniques available, each with its own approach. The
choice of algorithm depends on the specific dataset and application. Here are some
common classification algorithms and techniques [6], [12] , [19], [20] as shown in
Figure 1.

4
BIO Web of Conferences 97, 00133 (2024) https://doi.org/10.1051/bioconf/20249700133
ISCKU 2024

Fig .1. SL classification algorithms [15]

4.1 Zero R
Zero R or Zero Rule is the simplest classification method that relies solely on
the target data and disregards all other predictors. The Zero R classifier makes
predictions based on the majority category labels. Although Zero R lacks
predictability power, it is useful for establishing a baseline performance as a standard
for other classification methods [8], [18]. Zero R has advantage that it is provides
standard for other classification methods, and disadvantage is that it depends only on
target data.

4.2 One R
One R, also referred to as One Rule, is a simple classification algorithm that
generates one rule for each predictor in the data but is not highly accurate. It selects
the best predictor from a frequency table to predict the target, based on the smallest
total error using the one rule algorithm, and it is slightly less accurate than state-of-
the-art classification algorithms, but it can be useful for establishing a baseline
performance as a standard for comparison [19]. One R has several advantages in the
field of state-of-the-art classification, but it also has certain disadvantages.

4.3 Bayesian Algorithms

In Machine Learning, uncertainty can be quantified and managed using statistical
approaches, and Bayesian algorithms are one such approach based on probability
theory. These algorithms explicitly use Bayes’ Theorem for various tasks, including
classification and regression. There are several popular Bayesian algorithms, such as
in Figure 2 [7], [13], [21].
Figure 2 Bayesian algorithms
Belief Network (BBN) or Bayesian Network (BN) is a graphical model that can
illustrate the probability relationships between variables, such as symptoms and
diseases. It can be utilized to determine the probability of certain diseases based on
their corresponding symptoms. As a type of statistical learning algorithm, BBN is
commonly employed for probabilistic inference and decision making in various fields
[7], [13], [21].

5
BIO Web of Conferences 97, 00133 (2024) https://doi.org/10.1051/bioconf/20249700133
ISCKU 2024

Fig .2. Bayesian algorithms

In Bayesian Learning, practitioners select a prior probability distribution, which
is then systematically updated to derive a posterior distribution, which can be used as
the prior with new observations. The method is effective in handling incomplete
datasets and preventing over-fitting of data without requiring the removal of
contradictions. Bayesian Learning has practical applications in fields such as medical
diagnosis and disaster victim identification [22].
Naive Bayes (NB) is a machine learning algorithm based on Bayes' Theorem as
in Equation 1 and uses conditional probability to determine the likelihood of an object
belonging to a particular class.
x
P( ) P (c )
c c
P( )
x P( x) (1)
c x
Where P ( ) is the posterior probability, P ( ) is the likelihood, P(c) is the
x c
class prior probability, and P( x) is the predictor prior probability.
It is called "naive" because it operates under the assumption that the features used
for classification are independent of each other, which means that another variable
has no information about changes in any variable [22], , [23], [24].
NB is used for clustering and classification as in Figure 3 [14], [25], assigning a
posterior probability to a class based on its prior probability and likelihood for the
given training data. Although it violates the independence assumption, NB's
performance is competitive with most state-of-the-art classifiers and can be used in
binary and multiclass classification [8], [18], [20]. However, the NB classifier's slow
performance may occur as it rescans the entire dataset for each new classification
operation.

6
BIO Web of Conferences 97, 00133 (2024) https://doi.org/10.1051/bioconf/20249700133
ISCKU 2024

Fig .3. Naive Bayes [20]

The Naive Bayes algorithm is effective in real-world scenarios and can be used
to create reliable prediction models and accurately categorize noisy data. Even though
it requires only a small amount of statistical training to estimate necessary parameters,
its overall performance may suffer due to making too many assumptions about
function independence [10], [22]. Naive Bayes algorithm has several advantages and
disadvantages as illustrated in Table 1.
Table 1. Advantages and disadvantages of Naive Bayes [18], [22], [26], [27]
Advantages Disadvantages
It requires less training data, is easy to One of the problems of Naive Bayes is
implement, can handle continuous and known as the "Zero Conditional Probability
discrete data, is robust to noisy features, Problem," as it can remove all information
can make probabilistic predictions, is not from other probabilities, but this can be
sensitive to irrelevant features, has good addressed with the Laplacian Correction
performance, requires short computational technique. Naive Bayes assumes that all
time for training, can handle binary and class features are independent, which is
multi-class classification problems, is often not true in real-life data sets. NB
extremely fast compared to other models can be too simplistic, and models
classifiers, improves classification that are properly trained and tuned can
performance by removing irrelevant often outperform them. Directly applying
features, and its main strength is efficiency Naive Bayes to continuous variables (like
as training and classification can be done time) can be difficult, although "bucketing"
with one pass over the data. can be used to address this issue, it's not
100% correct. Real-world data may not
adhere to the conditional independence
assumption, leading to decreased
effectiveness of NB. When features are
highly correlated, NB performs poorly, and
the classifier is also sensitive to how input
data is prepared. To achieve good results, a
large number of records are needed, and
NB is not a good estimator compared to
other classifiers.
Naive Bayes is applicable in various areas including text analysis where it can
categorize words or phrases into pre-set groups or not, document classification, spam
filtering, sentiment analysis, recommendation systems, and forecasting the
progression of cancer relapse after radiotherapy [6], [11], [22].

7
BIO Web of Conferences 97, 00133 (2024) https://doi.org/10.1051/bioconf/20249700133
ISCKU 2024

4.4 Decision Tree

Decision trees (DT) or Classification trees are a predictive modeling
methodology used in statistics, data mining, is a popular and effective tool and are
among the earliest and prominent SL algorithms that is perfect for classification tasks
[6], [10], [23]. They are graphical representations as seen in Figure 4 [10] of a
problem or decision based on specific criteria, consisting of internal and external
nodes connected by y branches.

Fig .4. Decision tree [28]

An internal node evaluates a decision function to determine which child node to
visit next, while leaves indicate class labels and branches represent feature
combinations that lead to those class labels [7], [8].
DT mechanism is transparent easy to interpret and quick to learn and easy to
follow, and are a common component to many medical diagnostic protocols as e.g.
[8], [28], the process of constructing a decision tree involves breaking down the
dataset into smaller components until a tree with decision nodes and leaf nodes is
produced. Leaf nodes indicate a classification or decision. The root node in the tree
corresponds to the best predictor from the given datasets [18]. DT classifiers are used
to classify data by sorting them in a top-down approach from the root node to the leaf
node [14]. To reduce size and overfitting of the data, decision trees are pruned to
produce a classification tree which is then used for classification purposes [8]. Post-
pruning techniques are employed to evaluate the performance of decision trees as
they are pruned using a validation set [21]. The classification is based on the equally
exhaustive and mutually exclusive "if-then-else" situation [11]. The most popular
decision tree algorithms are shown in Figure 5.

Fig .5. Decision tree algorithms [13].

8
BIO Web of Conferences 97, 00133 (2024) https://doi.org/10.1051/bioconf/20249700133
ISCKU 2024

Decision Tree Algorithm has several advantages and disadvantages as illustrated

in Table 2.
Table 2 Advantages and disadvantages of DT [11], [18], [22], [26], [27]
Advantages Disadvantages
Simplicity, speed, scalability, accuracy, Firstly, they are prone to overfitting,
efficient memory usage, and the ability to resulting in complex decision rules that do
handle noisy data. It is a non-parametric not generalize well to new data due to a
algorithm, which means it does not require lack of inherent mechanism to stop
a lot of assumptions. Decision trees can splitting the data, creating a high variance
create complex decision boundaries, algorithm. Secondly, decision trees require
allowing them to easily solve non-linear a large volume of data to produce accurate
problems. They are highly versatile and can results. Additionally, decision trees can
perform multiple roles apart from the lead to large and complex tree structures,
standard predictions. making them difficult to interpret when the
Decision trees can perform incremental dataset has many entries. Moreover,
learning, use various measures such as decision trees can have complex
Entropy, Gini index, and Information gain representations for some concepts due to
to determine the best split attribute, and replication problems, and they may be
handle data with errors or missing values. prone to sampling error, leading to locally
They can handle both categorical and optimal solutions rather than globally
quantitative values and can fill in missing optimal solutions.
attribute values with the most probable The final decision in a decision tree can
value. The algorithm's high performance is depend on relatively few terms, and
due to the efficiency of its tree traversal decision trees can be highly time-
algorithm, and data preparation is easy. consuming to train, especially when
Additionally, decision trees explicitly dealing with multiple continuous
outline all possible alternatives while independent variables. In cases of
tracing each alternative. Therefore, imbalanced class datasets, the decision tree
decision trees are computationally model can become biased towards the
inexpensive, visually clear, and suitable for majority class, requiring adjustments to the
both regression and classification rows and columns of the dataset.
problems. Furthermore, decision trees use
optimization algorithms that look for pure
nodes at every level, without considering
the impact of recent decisions on the next
few stages of splitting, making them greedy
algorithms. Decision trees are also unstable
and high variance models, and some
changes in the data can significantly affect
the predictions produced by the model.
Finally, certain types of decision tree
algorithms may have limited performance
in solving regression problems, as they can
lose essential information and become
computationally expensive when dealing
with numerical dependent variables.
While decision trees offer both benefits and drawbacks, a notable constraint is
that they cannot be used over long periods, and are highly susceptible to data drifts.
DT have various practical applications such as exploring data, recognizing patterns,
pricing options in finance, and identifying disease and risk threats [16], [22].

9
BIO Web of Conferences 97, 00133 (2024) https://doi.org/10.1051/bioconf/20249700133
ISCKU 2024

4.5 Ensemble Algorithms

Ensemble algorithms are a machine learning technique that combines multiple
base models to create an optimal predictive model. By combining insights from
various learning models, ensemble methods help minimize errors caused by noise,
variance, and bias in the data. These models consist of multiple weak learner models
that are independently trained and whose predictions are merged in some way to
create an overall prediction. Ensemble methods divide the training data into subsets
for which independent learning models are constructed and then combined to form a
correct hypothesis as illustrated in Figure 6 [13] , [25], [29].

Fig .6. Ensemble algorithms [29]

Some popular examples of ensemble algorithms as illustrated in Figure 7.

Fig .7. Some popular ensemble algorithms [13] , [25], [29]

Random Forest (RF) or random decision is an ensemble learning method and
supervised machine learning algorithm widely used by data scientists for
classification and regression problems as in Figure 8.

Fig .8. Random forest [30]

10
BIO Web of Conferences 97, 00133 (2024) https://doi.org/10.1051/bioconf/20249700133
ISCKU 2024

The RF algorithm is an extension of the decision tree approach, where multiple

decision trees are constructed with the training data, similar to a forest with many
trees. The new data is then fitted within one of the trees, resulting in a "random
forest," which is used to determine the majority vote for classification or the average
for regression [6]. While Random Forests generally outperform decision trees, they
have lower accuracy than gradient boosted trees. Moreover, the performance of
Random Forests can be influenced by the characteristics of the data [17]. RF
algorithm has several advantages, including its ability to handle larger datasets,
stability, and robustness against noise and outliers. It can also solve both regression
and classification problems with higher accuracy compared to decision trees due to a
reduction in overfitting. And has certain disadvantages, including a longer training
period compared to other machine learning algorithms, high complexity that demands
more resources and computational time, and a slow real-time prediction [16], [31].
RF has a wide range of applications in various domains, including real-life
scenarios. For instance, it can be utilized in industries to assess the risk level of loan
applicants as either high or low risk. Additionally, RF can predict mechanical parts'
failure in automobile engines, social media share scores, and performance scores
[16].

4.6 Dimensionality Reduction (Discriminant Analysis)

Dimensionality reduction (DR) is a constructive way to address the issue of high-
dimensional data, which can lead to sparsity and computational problems. One
popular technique is discriminant analysis, which clusters objects based on their
similarity in features to classify data into different classes. Another approach is to
generate a set of primary variables through either feature deletion or extraction. This
reduces the size of the feature set and the number of dimensions. Principal component
analysis (PCA) is a common method for dimensionality reduction, which involves
converting higher-dimensional data into a smaller space while preserving all original
variables. For example, 3D data can be reduced to a smaller space such as 2D [7].
Dimensionality reduction algorithms are effective solutions to the problem of
curse of dimensionality, which arises when the number of dimensions increases and
the available data become sparse. This sparsity poses a challenge for methods that
require statistical significance, as obtaining a statistically sound and reliable result
often requires a growing amount of data to support the result exponentially with the
dimensionality. The study of dimensionality reduction methods aims to remove
irrelevant and redundant data to reduce the computational costs, and improve data
quality for efficient organization strategies. These algorithms seek and exploit the
inherent structure in the data in an unsupervised manner, similar to clustering
methods, and can be adapted for use in classification and regression [13]. Figure 9
shows the different dimensionality y reduction algorithms.
g

11
BIO Web of Conferences 97, 00133 (2024) https://doi.org/10.1051/bioconf/20249700133
ISCKU 2024

Fig .9. DR Algorithms [13]

Linear Discriminant Analysis (LDA) is a linear model commonly used in
dimensionality reduction for supervised classification problems in machine learning,
specifically in pattern classification problems that involve more than two classes.
LDA is also known as Normal Discriminant Analysis (NDA) or Discriminant
Function Analysis (DFA), and is a generalized form of Fisher’s Linear
Discriminant (FLD). The primary purpose of LDA is to separate samples of distinct
groups by transforming the data to a different space that is optimal for distinguishing
between the classes. LDA can project features from higher dimensional spaces into
lower dimensional ones to reduce resources and dimensional costs [32]. The
covariance matrix method is used for analysis, and LDA relies on the concept of
linear combinations of variables to best separate two classes (targets) [18].

4.7 Logistic regression

Logistic regression (LR) or Logit model, it is a probabilistic-based statistical
powerful method that is commonly used in supervised machine learning to solve
classification issues, it is a type of regression used for prediction, it can be considered
as an extension of linear regression. However, instead of producing continuous
outcomes, it produces a dichotomous outcome that represents the occurrence or non-
occurrence of an event, which is a popular tool for applied statistics and discrete data
analysis.
LR is a powerful method for solving classification problems by determining the
boundary between classes, where the outcome lies between 0 and 1, and a threshold
needs to be assigned to differentiate between two classes [23]. The binary
classification assigns a probability value to an input instance, where values higher
than 0.50 classify it as ‘class A’; otherwise, ‘class B’ as in Figure 10.

Fig .10. Logistic regression [31]

LR can be used for both numerical and categorical data and estimates the
probability of a new instance belonging to a certain class using a logistic
characteristic called the sigmoid function, mathematically defined in Equation 2
1
g ( z)
1  e z (2)
LR can be applied to both regression and classification problems, but its most
common application is classification [10], [21]. The goal of logistic regression is to
find the best-fitting relationship between the dependent variable, also referred to as
the target variable, and a set of independent variables, called the predictors. In this

12
BIO Web of Conferences 97, 00133 (2024) https://doi.org/10.1051/bioconf/20249700133
ISCKU 2024

context, the dependent variable is the variable of interest that we aim to predict, and
the independent variables are the factors that influence the outcome [16], [18]. LR
can also be generalized to model a categorical variable with more than two values,
known as the multinomial logistic regression. Logistic regression has the following
advantages and disadvantages as in Table 3.
Table 3 Advantages and disadvantages of LR [16], [22]
Advantages Disadvantages
Its advantages include simplicity of Logistic Regression has some limitations,
implementation, computational efficiency, including its inability to solve non-linear
and ease of regularization. No scaling is problems due to its linear decision surface,
required for input features. It is not affected and being prone to overfitting.
by small noise in the data and Additionally, this algorithm requires that
multicollinearity. As the output of Logistic all independent variables be identified to
Regression is a probability score, it is work well. Another major disadvantage of
required to specify customized performance the logistic regression algorithm is that it is
metrics so as to obtain a cutoff which can only suitable for binary predicted variables
be used to do the classification of the target. and assumes that the data is free of missing
Logistic regression is specifically designed values and that the predictors are
for classification problems, and it helps independent of each other.
understand how independent variables
influence the outcome of the dependent
variable. It is also the simplest algorithm
that does not require high computational
power and less prone to overfitting.
Additionally, the algorithm is precise and
can be updated easily to reflect new data.
LR is widely applied in various fields, such as risk identification for diseases,
word classification, weather prediction, and voting applications [16]. It is also used
for predicting the risk of developing a disease, cancer diagnosis, and engineering
applications such as predicting the probability of failure of a process, system, or
product [22].
Types of Logistic Regression
There are three different types of Logistic Regression algorithms:
Binary Logistic Regression: It has only two possible outcomes. For example,
yes or no.
Multinomial Logistic Regression: It has three or more nominal categories. For
example, cat, dog, elephant.
Ordinal Logistic Regression: It has three or more ordinal categories, ordinal
meaning that the categories will be in an order. For example, user ratings (1-5) [16].

4.8 K-Nearest Neighbours

K-Nearest Neighbours (KNN) is a machine learning algorithm that compares the
similarity metrics (such as Euclidean distance) of new data points to those of existing
data points to classify them, as in Figure 11.

13
BIO Web of Conferences 97, 00133 (2024) https://doi.org/10.1051/bioconf/20249700133
ISCKU 2024

Fig .11. K-nearest neighbours [33]

KNN works by having the closest neighbours of each data point vote on its
categorization using a simple majority. Its accuracy depends on the quality of data,
and it is robust against noise. However, the main challenge in using KNN is choosing
the appropriate number of neighbours. KNN is widely used in statistical estimation
and pattern recognition based on proximity relations between objects [10], [17], [18],
[23]. It is a non-parametric algorithm that does not assume any underlying data
distribution. KNN is a classification algorithm that uses a database with data points
grouped into several classes and classifies the sample data point given to it [22].
When K-NN is used in classification, you calculate to place data within the
category of its nearest neighbour. If k = 1, then it would be placed in the class nearest
1. K is classified by a plurality poll of its neighbours [6]. Table 4 provide a summary
of the advantages and disadvantages of KNN algorithm.
Table 4 Advantages and disadvantages of KNN [16], [18], [22]
Advantages Disadvantages
Its simplicity, robustness to noisy training The relatively high cost of classifying
data, and efficiency even with large unknown records and the requirement for
datasets. Building the model is inexpensive distance computation of k-nearest
and it is a highly flexible classification neighbors. As the size of the training set
scheme that is suitable for multi-modal increases, the algorithm becomes
classes. KNN can sometimes be the best computationally intensive, and accuracy can
method, and it is versatile and useful for be degraded by noisy or irrelevant features.
both regression and classification tasks. Furthermore, KNN is a lazy learner that
does not perform any generalization on the
training data. Additionally, higher. Another
disadvantage is that determining the value
of K is not always straightforward, and the
algorithm requires high memory since it
needs to store all of the training data. In
some cases, the prediction stage might also
be slow when dealing with large data.
KNN algorithm has various applications, including industrial tasks that involve
searching for similar tasks compared to others, handwriting detection and image
recognition, video recognition and stock analysis. It is also used in recommendation
systems, medical diagnosis of diseases with similar symptoms, credit rating based on
feature similarity, and forecasting votes for different political parties. KNN is utilized
in the analysis by financial institutions before sanctioning loans [11], [16], [22].

4.9 Support Vector Machine

Vapnik proposed statistical learning theory based machine learning method which
is known as Support vector machine (SVM). SVMs are the newest and most widely

14
BIO Web of Conferences 97, 00133 (2024) https://doi.org/10.1051/bioconf/20249700133
ISCKU 2024

used state-of-the-art supervised machine learning technique, and convenient

technique for solving problems related to classification of data and learning and
prediction [21], [24], [25], [34]. SVMs explain the relationship between dependent
and independent variables by finding a maximized hyperplane that acts as a margin
between two classes of data [17], [18].
This margin is drawn in such a way that the distance between the margin and the
classes is maximized in an n-dimensional space that helps in doing the classification
of the data points [14], [21], and classification error is minimized [17], [25]. Support
vectors are the data points that are closest to the decision surface [34]. Figure 12
provides a simplified illustration of an SVM classifier.

Fig .12. Support vector machine [34]

The objects may or may not be linearly separable in which case complex
mathematical functions called kernels are needed to separate the objects which are
members of different classes. SVM aims at correctly classifying the objects based on
examples in the training data set [22]. SVM algorithms initially designed for binary
classification, but later the approach to solve multi-class problems were also
developed by using a set of hyperplanes. SVM is approach that is useful to analyze
the dataset and divide it into classes. Following are the advantages and disadvantages
of SVM are shown in the Table 5.
Table 5 Advantages and disadvantages of SVM [16], [22], [34]
Advantages Disadvantages
SVMs are advantageous in dealing with a SVM algorithm has some limitations in
wide variety of classification problems, dealing with large datasets due to the
including high dimensional and non- increase in training time and difficulty in
linearly separable problems. As finding an appropriate kernel function.
generalization is adopted in SVM so there SVM also struggles with noisy datasets and
is less probability of overfitting, and does not provide probability estimates,
another advantage of SVM is its memory while understanding the final SVM model
efficiency. It can handle both semi can be challenging. Moreover, a significant
structured and structured data. Support drawback of SVM is the requirement to set
vector machine algorithms work well when multiple key parameters correctly to
there is a separation between the classes in achieve excellent classification results.
the data.
The practical use cases of the SVM algorithm: Business applications for
comparing the performance of a stock over a period of time. Providing investment
suggestions, and classification of applications requiring accuracy and efficiency.
Additionally, SVM is applied in several fields such as cancer diagnosis, fraud

15
BIO Web of Conferences 97, 00133 (2024) https://doi.org/10.1051/bioconf/20249700133
ISCKU 2024

detection in credit cards, handwriting recognition, face detection, and text

classification [16], [22].
Certain algorithms, such as Logistic Regression and Support Vector Machines,
are designed specifically for binary classification and cannot handle more than two
classes. However, these algorithms can be adapted for multi-class problems by using
either strategy:
One-vs-Rest: Fit one binary classification model for each class vs. all other
classes, or
One-vs-One: Fit one binary classification model for each pair of classes.
On the other hand, standard classification algorithms used for binary or multi-
class classification cannot be directly applied to multi-label classification. Instead,
specialized versions of these algorithms, such as Multi-label Decision Trees, Multi-
label Random Forests, and Multi-label Gradient Boosting, must be used [12].

4.10 Artificial Neural Networks

The human brain can be seen as a highly complex, nonlinear, and parallel
computer, capable of organizing its structural units (neurons) to perform different
tasks, such as pattern recognition, perception, and motor control [35].
Artificial neural networks (ANNs), usually simply called neural networks (NNs)
or neural nets, also known as connectionist systems. The term "Artificial neural
network" refers to a sub-field of artificial intelligence similar to the brain, so that
computers will have an option to understand things and make decisions in a human-
like manner. Artificial neural networks are a set of machine learning algorithms (a
computational network) that use supervised learning (or unsupervised learning) and
are inspired by the functioning of the biological neural networks. In the biological
brain, neurons are connected to each other through multiple axon junctions forming
g
a graph like architecture [7], [13], [36] in Figure 13.

Fig .13. The typical diagram of biological NN [37]

Artificial Neural Network primarily consists of three layers:
Input Layer: As the name suggests, it accepts inputs in several different formats
provided by the programmer.
Hidden Layer: The hidden layer presents in-between input and output layers. It
performs all the calculations to find hidden features and patterns.
Output Layer: The input goes through a series of transformations using the hidden
layer, which finally results in output that is conveyed using this layer [38] as in Figure
14.

16
BIO Web of Conferences 97, 00133 (2024) https://doi.org/10.1051/bioconf/20249700133
ISCKU 2024

Fig .14. The basic components of the artificial network two layers [36]
Neural Networks have the capability to handle multiple regression and
classification tasks simultaneously [21]. However, using ANN can be challenging
due to the time required for training on complex data, and their black box nature
where users cannot interfere with the final decision-making process [17]. The concept
of ANN was introduced by McCulloch and Pitts in 1943, with the aim of simulating
the functions and structure of living beings' nervous systems [38] as shown in Figure
15, serving as a base model for Rosenblat's Perceptron [35]

Fig .15. McCulloch-Pitts artificial neuron [35]

There are various types of ANNs, with the most popular being:
Perceptron: The Perceptron, studied by F. Rosenblatt, is a significant
advancement in automated learning for pattern recognition [35]. It serves as an
algorithm for binary classification ( "yes" or "no"; 1 or 0; etc), separating input data
into distinct classes by producing a separating line. The Perceptron is widely used in
machine learning tasks such as gender determination, disease risk assessment, and
virus detection. Multi-Layered Perceptron (MLP) is the most often used, as one of
the simplest networks, and is represented in Figure 16.

Fig .16. Multi-layered perceptron [39]

17
BIO Web of Conferences 97, 00133 (2024) https://doi.org/10.1051/bioconf/20249700133
ISCKU 2024

Within the realm of popular algorithms, there are also:

Hopfield Network (HN), Back-Propagation, and Radial Basis Function
Network (RBFN) [13].
Advantages ANNs, it has a high tolerance to noisy data and able to classify
untrained patterns, it performs better with continuous-valued inputs and outputs. The
disadvantage with the artificial neural networks is that it has poor interpretation
compared to other models. (ANNs) have found extensive applications in various
fields, including medicine, transportation, optimization, and even quantum physics.
Some notable use cases of ANNs include handwriting analysis, colorization of black
and white images, computer vision processes, and captioning photos based on facial
features [16], [38].

4.11 Deep Learning Algorithms

Deep Learning Algorithms (DLA) are advanced machine learning (Learning can
be supervised, semi-supervised or unsupervised) methods that utilize Artificial
Neural Network (ANN) structures and leverage available computing resources to
construct larger and more complex networks. The "deep" in deep learning refers to
the use of multiple layers, typically more than three, with numerous nodes in each
layer, making them suitable for large datasets [39]. DL is a branch of Artificial
Intelligence that mimics the human brain's information processing and decision-
making abilities, incorporating techniques like statistics and predictive modeling. An
advantage of deep learning is its ability to develop its own feature set without
supervision [14]. In DL, classification problems are addressed through the training of
classification models. These models are trained using labeled objects, allowing them
to learn and recognize common features within a particular class. Once trained, the
models are tested on separate data, where only the object to be classified is provided
without its label. The classification model then predicts the label for the object, and
the accuracy of the model is evaluated based on the correct predictions of labels. The
most popular deep learning algorithms are:
Feed-Forward Neural Network (FFNN) is a type of neural network that operates
in a unidirectional manner, without loops or feedback paths. It comprises input,
output, and hidden layers, as shown in Figure 17 [18].

Fig .17. FFNN [40]

Feed-Back Neural Network (FBNN), also called recurrent neural network
(RNN) is a type of neural network that features back propagation of feedback paths,
allowing signals to move in both forward and backward directions through repetitive
loops. This network structure permits all possible connections between neurons. Due
to the presence of these loops, the feedback network becomes a non-linear dynamic
system that undergoes continuous changes until it reaches a state of equilibrium, as
illustrated in Figure 18 [18].

18
BIO Web of Conferences 97, 00133 (2024) https://doi.org/10.1051/bioconf/20249700133
ISCKU 2024

Fig .18. FBNN [40]

Convolutional Neural Networks (CNNs): CNNs are regularized versions of
multilayer perceptrons. Multilayer perceptrons are usually fully connected networks,
that is, each neuron in one layer is connected to all neurons in the next layer. CNNs
use a mathematical operation called convolution in place of general matrix
multiplication in at least one of their layers. CNNs are a type of feedforward neural
network. They are designed to process data with a grid-like topology, and are able to
learn features and patterns in the data through the use of convolutional layers. The
neurons present in this network are have learnable weights and biases. Each neuron
receives some inputs, performs a dot product and optionally follows it with a non-
linearity [38]. CNNs derives its inspiration from the biological processes that occur
in the visual lobe specifically in the brain of living beings, and is considered a solution
to many computer vision problems in artificial intelligence, such as processing
images and videos [38]. Other most popular deep learning algorithms include:
Deep Boltzmann Machine (DBM), Deep Belief Networks (DBN), and Stacked
Auto-Encoders [13].

5 LEARNER IN CLASSIFICATION PROBLEMS

Classification models can learn in two different ways:
Lazy learners also known as instance-based learners, here the training data is
stored until test data is available to classify.The algorithm does not build a model
until it receives test data. However, due to the constant updating of data, these
algorithms quickly become outdated. Although they require less time for training,
these algorithms take longer to predict as they make local approximations based on
the most similar stored data.This approach is useful when dealing with large and
changing datasets with a smaller set of queried attributes. Lazy learning is easy to
maintain and can be applied to various problems.The k-nearest neighbor (KNN)
algorithm is an example of a lazy learner.
Eager Learners is a type of machine learning where, eager learners create a
classification layer prior to training and testing the dataset. They construct an explicit
description of the training function based on the provided data, committing to a single
hypothesis that covers the entire dataset. Eager learning takes longer to train the
dataset and less time to predict compared to the lazy learning system due to the
creation of a classification model. Decision trees, naive Bayes, and artificial neural
networks (ANN) are examples of eager learners [16], [20].

6 CONCLUSION

19
BIO Web of Conferences 97, 00133 (2024) https://doi.org/10.1051/bioconf/20249700133
ISCKU 2024

Machine learning has evolved into an indispensable component of individuals'

daily lives, particularly in the realm of supervised learning. This methodology has
witnessed a constant surge in popularity. Supervised learning, a pivotal branch of
machine learning, entails the training of models on labelled datasets and subsequently
evaluating them on unlabelled data. This process can be further categorized into
classification and regression tasks. Classification, in particular, holds significant
value as it serves as a robust technique for predicting information from either
categorical or numerical datasets. This technique plays a pivotal role in statistical
problem analysis.
In classification, the approach involves the utilization of categorized training data
to assign labels to corresponding datasets. Various methodologies exist for creating
class models. The choice of classification and classifiers depends on the relationship
between the dependent and independent variables, aiming to achieve optimal class
labelling. The algorithms encompass distinct data processing and feature engineering
techniques to facilitate the classification process. For any data scientist, a
comprehensive understanding of classification is of paramount importance.
Classification serves as a foundational concept underpinning numerous other
techniques and domains within machine learning.
Proficiency in classifying and recognizing specific data types empowers computer
scientists to broaden their knowledge and extend the applications in diverse machine
learning fields, including computer vision, natural language processing, deep
learning, predictive economic, market, and weather models, among others.
Nevertheless, evaluating the superiority of one classification technique over
another presents a formidable challenge, given that each approach exhibits its own
merits, demerits, and implementation intricacies, which often vary based on the
problem domain of the user. Despite significant progress in the field of classification,
a pressing need exists for comprehensive attention from the research community to
address emerging challenges, especially concerning the handling and classification
of Big Data.
Consequently, it is of paramount importance to attain a comprehensive
understanding of the definitions, advantages, and limitations associated with popular
supervised machine learning algorithms utilized for classification. This review has
provided a brief overview of these algorithms, highlighting their definitions,
advantages, and limitations.
The principal findings underscore the significance of classification techniques in
categorizing data based on its distinctive features and attributes, thereby enabling
predictive modelling and effective decision-making. It is imperative to consider the
unique characteristics of classification techniques, including computational
complexity, management of high-dimensional data, sensitivity to noise, and
interpretability.
Moreover, an awareness of their applications across a wide spectrum of domains
such as healthcare, finance, image recognition, and text analysis are vital when
selecting the most suitable technique. By deepening our comprehension of
classification techniques and exploring innovative approaches, we can propel
advancements in the realm of machine learning and cultivate its practical
applications.
To summarize, a profound understanding of classification techniques is critical
for well-informed decision-making, enhanced accuracy and effective utilization
across diverse domains. Addressing challenges, embracing novel approaches, and

20
BIO Web of Conferences 97, 00133 (2024) https://doi.org/10.1051/bioconf/20249700133
ISCKU 2024

uncovering applications in emerging fields are instrumental in contributing to the

evolution of machine learning and ultimately benefiting society.

ACKNOWLEDGEMENT
I would like to thank the reviewers for providing useful suggestions, allowing for the improved
presentation of this paper. And I would also like to extend my thanks and gratitude to my
honorable supervisor for her continuous support and dedication to tirelessly providing
scientific advice to make this work a success.

References

1. "Indian Institute of Technology Madrras," CS6464: Concepts In

Statistical Learning Theory, January - May 2023. [Online]. Available:
http://www.cse.iitm.ac.in/~vplab/statistical_learning_theory.html.
2. J. P. Mueller and L. Massaron, Machine Learning For Dummies,
Hoboken, New Jersey: John Wiley & Sons, Inc, 2016.
3. "javatpoint.com," Machine Learning, [Online]. Available:
https://www.javatpoint.com/machine-learning. [Accessed 6 feb 2023].
4. O. Theobald, Machine Learning For Absolute Beginners, London, UK:
Scatterplot press, 2017.
5. S. Theodoridis, Machine Learning A Bayesian and Optimization
Perspective, Elsevier Ltd, 2020.
6. R. Wolff, " MonkeyLearn Blog," 5 Types of Classification Algorithms in
Machine Learning, 26 August 2020. [Online]. Available:
https://monkeylearn.com/blog/classification-algorithms/.
7. R. Praba, G. Darshan, K. T. Roshanraj. and P. B. Surya Prakash, "Study
On Machine Learning Algorithms," International Journal of Scientific
Research in Computer Science, Engineering and Information Technology,
vol. 7, no. 4, pp. 67-72, 2021.
8. C. Nasa and Suman, "Evaluation of Different Classification Techniques
for WEB Data," International Journal of Computer Applications, vol. 52,
no. 9, p. 0975 – 8887, 2012.
9. N. . H. Ali, M. Emaduldeen and A. E. Ali, "Learning Evolution: a Survey,"
Iraqi Journal of Science, vol. 62, no. 12, pp. 4978-4987, 2021.
10. A. Ram and Meenakshi, "Short Review on Machine Learning andits
Application," SPECIALUSIS UGDYMAS / SPECIAL EDUCATION, vol.
1, no. 43, pp. 9894-9902, 2022.
11. S. Bansal, "ANALYTIXLABS," MACHINE LEARNING What is
Classification Algorithm in Machine Learning? With Examples, 25
JANUARY 2023. [Online]. Available:
https://www.analytixlabs.co.in/blog/classification-in-machine-learning/.

21
BIO Web of Conferences 97, 00133 (2024) https://doi.org/10.1051/bioconf/20249700133
ISCKU 2024

12. J. Brownlee, "Machine Learning Mastery," 4 Types of Classification

Tasks in Machine Learning, 19 Aug 2020. [Online]. Available:
https://machinelearningmastery.com/types-of-classification-in-machine-
learning/.
13. S. N. dhage and C. K. Raina, "A review on Machine Learning
Techniques," International Journal on Recent and Innovation Trends in
Computing and Communication, vol. 4, no. 3, pp. 395 - 399, 2016.
14. V. Rastogi, S. Satija, P. K. Sharma and S. Singh, "MACHINE LEARNING
ALGORITHMS:OVERVIEW," International Journal of Advanced
Research in Engineering and Technology (IJARET), vol. 11, no. 9, pp.
512-517, 2020.
15. P. Ariwala, "Maruti Techlabs," 9 Real-World Problems that can be
Solved by Machine Learning, 16 Oct 2023. [Online]. Available:
https://marutitech.com/problems-solved-machine-learning/.
16. M. Waseem, "Edureka Blog," How To Implement Classification In
Machine Learning?, 2 Aug 2023. [Online]. Available:
https://www.edureka.co/blog/classification-in-machine-learning/.
17. M. Yuval, B. Yaman and Ö. Tosun, "Classification Comparison of
Machine Learning Algorithms Using Two Independent CAD Datasets,"
Mathematics , vol. 10, no. 3, p. 311, 2022.
18. R. V. K. Reddy and U. R. Babu, "A Review on Classification Techniques
in Machine Learning," International Journal of advance research in
science and engineering, vol. 7, no. 3, pp. 40-47, 2018.
19. D. Michie, D. J. Spiegelhalter and C. C. Taylor, Machine Learning,
Neural and Statistical Classification, 1994.
20. PI.EXCHANGE, "PI.EXCHANGE," Understanding Classification in
Machine Learning , 16 Jan 2023. [Online]. Available:
https://www.pi.exchange/blog/understanding-classification-in-machine-
learning.
21. F. Y. Osisanwo , J. T. Akinsola , J. O. Hinmikaiye , O. Olakanmi and J.
Akinjobi , "Supervised Machine Learning Algorithms: Classification and
Comparison," International Journal of Computer Trends and Technology
(IJCTT), vol. 48, no. 3, pp. 128-138, 2017.
22. S. Ray, "A Quick Review of Machine Learning Algorithms," in 2019
International Conference on Machine Learning, Big Data, Cloud and
Parallel Computing (Com-IT-Con), India, 14th -16th Feb 2019, India,
2019.
23. M. A. Kashmoola, M. K. Ahmed and N. . Y. A. Alsaleem, "Network Traffic
Prediction Based on Boosting Learning," Iraqi Journal of Science, vol.
63, no. 9, pp. 4047-4056, 2022.

22
BIO Web of Conferences 97, 00133 (2024) https://doi.org/10.1051/bioconf/20249700133
ISCKU 2024

24. F. H. Fadel and S. F. Behadili, "A Comparative Study for Supervised

Learning Algorithms to Analyze Sentiment Tweets," Iraqi Journal of
Science, vol. 63, no. 6, pp. 2712-2724, 2022.
25. C. Shelke, "Machine Leaning and it’s Various Algorithms- A Study,"
Journal of Emerging Technologies and Innovative Research (JETIR), vol.
8, no. 5, pp. 258-265, 2021.
26. S. D. Jadhav and H. P. Channe, "Comparative Study of K-NN, Naive
Bayes and Decision Tree Classification Techniques," International
Journal of Science and Research (IJSR), vol. 5, no. 1, pp. 1842-1845,
2016.
27. N. Kalcheva, M. Todorova and G. Marinova, "NAIVE BAYES
CLASSIFIER, DECISION TREE AND ADABOOST ENSEMBLE
ALGORITHM – ADVANTAGES AND DISADVANTAGES," in 6th
International Scientific Conference ERAZ - Knowledge Based Sustainable
Development, Belgrade, Serbia, 2020.
28. T. 365 Team, "The 365 Team," Introduction to Decision Trees: Why
Should You Use Them?, 12 Apr 2023. [Online]. Available:
https://365datascience.com/tutorials/machine-learning-tutorials/decision-
trees/.
29. ValianceSolutions, "data science central," Improving Predictions with
Ensemble Model, 8 Oct 2016. [Online]. Available:
https://www.datasciencecentral.com/improving-predictions-with-
ensemble-model/.
30. spotfire, "spotfire," What is a random forest?, [Online]. Available:
https://www.spotfire.com/glossary/what-is-a-random-forest.
31. T. F. Beginner, "Tutorial For Beginner," Linear Regression vs Logistic
Regression in Machine Learning, [Online]. Available:
https://tutorialforbeginner.com/linear-regression-vs-logistic-regression-
in-machine-learning.
32. K. S. S. Kumar, "DOC493: Intelligent Data Analysis and Probabilistic
Inference Lecture 15.," Imperial College., 2015.
33. IBM, "IBM," What is the k-nearest neighbors algorithm? , [Online].
Available: https://www.ibm.com/topics/knn.
34. A. . A. Soofi and A. Awan, "Classification Techniques in Machine
Learning: Applications and Issues," Journal of Basic & Applied Sciences,
vol. 13, pp. 459-465, 2017.
35. W. Z. Cárdenas, M. Zumbado and T. Zelaya, "McCulloch-Pitts Artificial
Neuron and Rosenblatt's Perceptron: An abstract specification in Z,"
Revista Technology, vol. 5, pp. 16-29, 2020.
36. j. T. point, "java T point," Artificial Neural Network Tutorial, [Online].
Available: https://www.javatpoint.com/artificial-neural-network.

23
BIO Web of Conferences 97, 00133 (2024) https://doi.org/10.1051/bioconf/20249700133
ISCKU 2024

37. Wikipedia, "Wikipedia," Biological neuron models, [Online]. Available:

https://en.wikipedia.org/wiki/Biological_neuron_model.
38. J. D. Pineda-Jaramillo, "A review of Machine Learning (ML) algorithms
used for modeling travel mode choice," DYNA, vol. 86, no. 211, pp. 2-41,
2019.
39. J. Djuris, I. Kurcubic and S. Ibric, "Review of machine learning
algorithms application in pharmaceutical technology," Arhiv za
farmaciju , vol. 71, no. 4, pp. 302-317, 2021.
40. R. Quiza and J. P. Davim, "Computational methods and optimization," in
Machining of hard materials, Verlag, Springer, 2011, pp. 177-208.

24