NIET Journal of Engineering & Technology (NIETJET)

Volume 10, Issue summer 2022 ISSN: 2229-5828 (Print)

A detailed analysis of the supervised

machine Learning Algorithms
Kriti Bansal1, Preeti Gupta1
1
Student, Noida Institute of Engineering and Technology, Greater Noida

ABSTRACT: In the field of computer science known as "machine learning," a computer makes predictions about
the tasks it will perform next by examining the data that has been given to it. The computer can access data via
interacting with the environment or by using digitalized training sets. In contrast to static programming
algorithms, which require explicit human guidance, machine learning algorithms may learn from data and
generate predictions on their own. Various supervised and unsupervised strategies, including rule-based
techniques, logic-based techniques, instance-based techniques, and stochastic techniques, have been presented in
order to solve problems. Our paper's main goal is to present a comprehensive comparison of various cutting-edge
supervised machine learning techniques.

1. INTRODUCTION
Machine learning offers systems the capacity to learn automatically and improve over time without explicit
coding. Machine learning algorithms are helpful in situations where it is unfeasible to deploy explicitly written
algorithms with high performance. Giving some numbers as input and receiving an ordered list as an output makes
it straightforward to complete a task like sorting integers. Here, we know what to provide as input and the steps
to take to get the result we want. However, some tasks are difficult to understand, such as email filtering to separate
spam from valid communications. Here, we are aware of the required input and the form of the output, which is
true. Throughout this case, we are aware of the required input and that the output will take the form of true or
false, but the instructions that must be given to the programme in order for it to carry out these operations are
unclear. We use data to our advantage and give instructions to the machine to evaluate the data and interpret it
intelligently in such unusual scenarios where there is no set algorithm to accomplish success [1]. Concrete
technology is constantly being revised and improved since it is relatively inexpensive compared to other building
materials and is frequently employed in engineering constructions around the world. [1]. Concrete is in high
demand due to urbanization's quick and technologically advanced development [2], as it has numerous desirable
features such compressive strength, shape-ability, and environmental resistance [3]. The benefits of concrete are
also listed as including porosity, damage tolerance, fire resistance, durability, and acoustic insulation. [4].

2. Learning Strategies
Machine learning employs the following strategies:

2.1. Supervised learning:

A. Regression:
i. Linear Regression:
The outcome of linear regression is achieved by adding the inputs and multiplying them by a set of constants, to
put it simply.
It uses a straight line to establish a correlation between Y, a dependent variable, and X, which may be a number
of independent variables (regression line).
The general equation can be written as - Y = a + bX,
Where, Y – Dependent Variable, X – Explanatory Variable
ii. Support Vector Machine Regression:

34 | Page
Publisher: Noida Institute of Engineering & Technology,
19, Knowledge Park-II, Institutional Area, Greater Noida (UP), India.
 NIET Journal of Engineering & Technology (NIETJET)
Volume 10, Issue summer 2022 ISSN: 2229-5828 (Print)

If a specific training set is provided, such as (x1, y1), (xi, yi) X R, where X is the space of input patterns. In SV
regression, our objective is to find a fitting function f(x) with a deviation from the target (yi) gained for the relevant
training data set that is smaller than. The function ought to be rather flat. Or you may say that any error that is less
than is acceptable [12]. The linear equation (f) - f (x) = (w, x) + b where (•, •) is the dot product of X and w denotes
flatness in this instance. We must limit the norm to a minimum in order to be sure.

iii. Decision Tree Regression:

This regression process operates by dividing a dataset into smaller subsets, after which a corresponding decision
tree is incrementally created. Finally, a tree with leaf nodes and decision nodes is created. The root node of the
tree corresponds to the best predictor and is the topmost decision node [14]. The decision tree is constructed using
the ID3 (Iterative Dichotomiser) fundamental method. The decision tree is built by the ID3 method using Standard
Deviation Reduction (SDR).
Steps involved in SDR:
a) Initially, we calculate the standard deviation of the target.
b) After this, we divided the datasets based on the various attributes. The standard deviation before the split is
then reduced by the standard deviation for each branch that results. The SDR is this. S(T) - S(T,X) =
SDR(T,X) iii As the decision node, the property with the highest SDR should be chosen. iv. Based on the
values of the chosen attributes, the dataset needs to be partitioned. If the standard deviation is more than 0,
we divide the branch set again. When all of the data has been processed, the process continues to run in
recursion.

iv. Random forest:

It adds an extra layer of unpredictability to bagging. In contrast to a typical tree, a random forest divides each
node using the best predictor from a subset that is randomly chosen at that node. Its few parameters—the number
of trees in the forest and variables at each node in the random subset—make it a simple technique to utilize. [15].
The RF Regression algorithm steps are: i. Take an nt bootstrap sample based on the real data. ii. Every bootstrap
sample requires the construction of a regression tree with a few modifications: selecting the optimum variable
split and sampling a certain number of predictors at random from each node. iii. Cumulate the predictions of the
nt trees to estimate the most recent data (average for regression). LASSO, which stands for "least absolute
shrinkage and selection operator," is a shorthand for regression. It maintains the quality characteristics of ridge
regression and subset selection by reducing some coefficients and setting others to zero. In addition to reducing
variability and increasing precision in linear regression models, LASSO regression also penalises the total size of
the regression coefficients. [16]
Since the penalty function in this regression utilizes absolute values rather than squares, certain parameter
estimates are accurate to the nearest tenth. B. Identification: i. Reliable method to solve binary classification
problems is logistic regression. In order to estimate the likelihood of an outcome having just two values, one uses
logistic regression. The logistic function, an S-shaped curve that maps any real-valued number into a value
between 0 and 1, though never precisely at 0 and 1, is the cornerstone of logistic regression. [17].
v. K Nearest Neighbors:
Due to its ease of interpretation and quick calculation times, this approach is renowned for its simplicity. In
essence, it retains the existing examples and classifies new cases using homogeneity criteria like distance
functions. A majority of the object's neighbors vote to categories it, and the outcome is usually class integration.
After that, the object is assigned to the class with the K nearest neighbors that share the most similarities [19].
Categorical variables require the usage of Hamming distance. iii. Simple to set up and suitable for use with big
datasets of data, nave Bayesian is based on the probabilistic Bayes theorem and is based on a probabilistic model.
The nave Bayesian classifiers make the assumption that the value of one characteristic, given the class variables,
is independent of the value of any other characteristic.
P (B|A) is the probability of the predictor when the class is supplied, P (B) is the prior probability of the predictor,
and P (A) is the posterior probability of the class when the predictor is given (attribute).

35 | Page
Publisher: Noida Institute of Engineering & Technology,
19, Knowledge Park-II, Institutional Area, Greater Noida (UP), India.
 NIET Journal of Engineering & Technology (NIETJET)
Volume 10, Issue summer 2022 ISSN: 2229-5828 (Print)

vi. Decision Tree (DT) Classification:

Instead of using the Standard Deviation Reduction method to build a decision tree in the decision tree
classification, the ID3 algorithm uses entropy and information gain. Calculating the homogeneity of the sample
involves using entropy. Entropy must be zero in order for the sample to be completely homogeneous, which occurs
when the sample is divided into equal portions [21].
vii. Bagging
The arrangement of bagging in a way that can increase the stability and accuracy of the machine learning
algorithms used in regression and classification is known as bootstrap aggregating or bagging. It is typically
applied to lessen the differences between actual and expected results. Although bagging can be used with any sort
of method, it is most frequently used with decision tree approaches. It is also regarded as one of the model
averaging technique's special situations.
Bagging is a parallel ensemble machine material that, by giving supplemental data during the training phase,
explains the variance of projected models. In the new dataset, each element has an equal chance of showing up.
While changing the training set, predictive power cannot be increased. An optimal value for the decision tree with
bagging is modulated using 20 sub-models, and as a result, a solid output result can be obtained.

viii. Gradient boosting

The general consensus is that one of the effective methods for developing predictive models is gradient boosting.
It is an ensemble machine learning algorithm that is frequently used for classification and regression issues. It
creates a projected model out of a collection of weak predicted models, typically a decision tree. The resulting
technique is therefore referred to as a gradient boosting tree when the decision tree outputs the result as a weak
learner. The field of learning to rank can also benefit from the use of gradient boosting. It is also employed in data
analysis for high energy physics. The artificial neural network (ANN) algorithm has a network of neurons that
resembles the brain. The ANN, which serves as a model of the human brain, is essentially a collection of
interconnected units or nodes (sometimes referred to as artificial neurons). These neural networks pick up
information by processing examples. They create probability-weighted associations between the input and result
and are stored within the data structure of the net itself. They contain a known "input" and "result." Today, there
is a lot of interest in using ANNs in the field of civil engineering, particularly to forecast the mechanical
characteristics of concrete. This is because it can anticipate concrete's real strength results with a high degree of
precision.

3. Parameters used for evaluating tool performance

It is not appropriate to evaluate the success of the classifying algorithm solely by the accuracy value attained by
any classifier. The value for the instances categorised as belonging to their true class is what determines the
classifier's accuracy. It omits discussing other classifier-specific details like the relationship between data
attributes, the measurement of how accurately data instances are distributed across all possible classes, the number
of positive results out of all received positive results, and a number of other things.
These factors are also crucial for evaluating the effectiveness of any classifier, which has further aided in the
comparative analysis of particular tools used in this research. Some of the parameters that are used in the
evaluation process are listed below. Recall, Accuracy, and Confusion Matrix:
The value for the instances categorised as belonging to their true class is what determines the classifier's accuracy.
It does not introduce to the other specificities of the classifier like; the relation between the data attributes, the
measure of correct distribution of data instances to each and every class possible, the number of the positive
outcomes from the among all received positive outcomes, and several other. These factors are also crucial for
evaluating any classifier's performance, which has further aided in this research's comparative analysis of various
methods.
Some of the parameters utilised in the evaluation process are discussed in the sections that follow. I Accuracy,
confusion matrix, and recall: A confusion matrix is a type of table that indicates how many instances of data are
incorrectly classified and how many are correctly classified. This matrix, nXn, represents the number of classes
that were established for the data collection. Figure 4: Confusion Matrix From this confusion matrix, values are

36 | Page
Publisher: Noida Institute of Engineering & Technology,
19, Knowledge Park-II, Institutional Area, Greater Noida (UP), India.
 NIET Journal of Engineering & Technology (NIETJET)
Volume 10, Issue summer 2022 ISSN: 2229-5828 (Print)

derived for nearly all other parameters. The rows reflect the actual values or class labels to which the data object
truly belongs, while the columns indicate the predicted values by the classifier as shown in Fig 1.
A confusion matrix is a sort of table that counts the number of instances of data that are correctly and wrongly
categorised. The number of established classes is indicated by the matrix nXn. for the data collection.

Fig 1. Confusion Matrix

The values of practically all other parameters are derived from this confusion matrix. The rows represent the actual
values or class labels to which the data object actually belongs, while the columns represent the predicted values
by the classifier. The cell values are as follows:
 True negative: What percentage of the negative cases was accurately classified? It is calculated as:
TN/TN+FP
 False Positive: What percentage of negative instances was wrongly labelled as positive? It is calculated as:
FP/TN+FP
 False Negative: What percentage of positive cases was wrongly labelled as negative? It is calculated as:
FN/FN+TP
 True Positive or Recall: percentage of positively classified positive instances. Accuracy is calculated as
TP/FN+TP and represents the proportion of accurate predictions made by the classifier. It is calculated as:
TN+TP/TN+FP+FN+TP.
 Precision (Confidence): Precision is the percentage of positive outcomes that were anticipated to be positive
and ultimately turned out to be so. It is calculated as TP/FP+TP

4. Conclusions and Future Recommendations

This study offers details on supervised machine learning techniques used individually and collectively to estimate
the compressive strength of concrete at high temperatures. When compared to the actual result, the use of ML
approaches for concrete performance prediction demonstrates a high level of accuracy, making it a very useful
strategy. The average duration required to determine the strength of concrete is 28 days. In consequence, ML
algorithms contribute significantly to shortening this period of time and also substantially reduce the costs and
labour necessary to carry out experimental activity.
This study offers details on supervised machine learning techniques used individually and collectively to estimate
the compressive strength of concrete at high temperatures. When compared to the actual result, the use of ML
approaches for concrete performance prediction demonstrates a high level of accuracy, making it a very useful
strategy. The average duration required to determine the strength of concrete is 28 days. In consequence, ML
algorithms contribute significantly to shortening this period of time and also substantially reduce the costs and
labour necessary to carry out experimental activity.
However, the models can also be run with additional input factors, such as temperature and other relevant effects
Materials 2021, 14, 4222 13 of 19, including humidity, to get the desired output.

References:
[1] Domingos, P. “A few useful things to know about machine learning”, Communications of the ACM,
55(10),2012 pp.1.

37 | Page
Publisher: Noida Institute of Engineering & Technology,
19, Knowledge Park-II, Institutional Area, Greater Noida (UP), India.
 NIET Journal of Engineering & Technology (NIETJET)
Volume 10, Issue summer 2022 ISSN: 2229-5828 (Print)

[2] Mohri, M., Rostamizadeh, A. and Talwalker, A. “Foundations of machine learning”, Cambridge, MA: MIT
Press,2012.
[3] Nguyen, T. and Shirai, K. “Text Classification of Technical Papers Based on Text Segmentation”, Natural
Language Processing and Information Systems, 2013,pp.278-284.
[4] Deng, L. and Li, X. “Machine Learning Paradigms for Speech Recognition: An Overview”, IEEE
Transactions on Audio, Speech, and Language Processing, 21(5), 2013, pp.1060-1089.
[5] Siswanto, A., Nugroho, A. and Galinium, M. “Implementation of face recognition algorithm for biometrics
based time attendance system”, 2014 International Conference on ICT For Smart Society (ICISS).
[6] Chen, Z. and Huang, X. “End-to-end learning for lane keeping of self-driving cars”, 2017 IEEE Intelligent
Vehicles Symposium (IV).
[7] Yong, S., Hagenbuchner, M. and Tsoi, A. “Ranking Web Pages Using Machine Learning Approaches”, 2008
IEEE/WIC/ACM International Conference on Web Intelligence and Intelligent Agent Technology.
[8] Wei, Z., Qu, L., Jia, D., Zhou, W. and Kang, M. “Research on the collaborative filtering recommendation
algorithm in ubiquitous computing”, 2010 8th World Congress on Intelligent Control and Automation.
[9] Kononenko, I. “Machine learning for medical diagnosis: history, state of the art and perspective”, Artificial
Intelligence in Medicine, 23(1), 2011, pp.89-109.
[10] Jordan, M. “Statistical Machine Learning and Computational Biology”,IEEE International Conference on
Bioinformatics and Biomedicine (BIBM 2007).
[11] Thangavel, S., Bkaratki, P. and Sankar, A. “Student placement analyzer: A recommendation system using
machine learning”, 4th International Conference on Advanced Computing and Communication Systems
(ICACCS-2017).
[12] Byun, H. and Lee, S., “Applications of Support Vector Machines for Pattern Recognition: A Survey”. Pattern
Recognition with Support Vector Machines, 2002, pp.214-215.
[13] Support vector machine regression algorithm [Online],
http://chemeng.utoronto.ca/~datamining/dmc/support_vector_machine_reg.htm, last access 22.08.2017.
[14] Kotsiantis, S. “Decision trees: a recent overview. Artificial Intelligence Review”, 39(4), 2011, pp.262-267.
[15] Andy Liaw and Matthew Wiener “Classification and Regression by randomForest”, R News, ISSN 1609-
363, vol. 2/3, December 2002, pp. 18-22.
[16] Tibshirani, R.”Regression shrinkage and selection via the lasso: a retrospective”, Journal of the Royal
Statistical Society: Series B (Statistical Methodology), 73(3), 2011, pp.273-282.
[17] Brownlee, J. “Logistic Regression for Machine Learning - Machine Learning Mastery”, [online] Machine
Learning Mastery. Available at: http://machinelearningmastery.com/logistic-regression-for-machine-
learning/ [Accessed 12 Aug. 2017].
[18] The steepness of the curve of logistic regression [Online],
http://chemeng.utoronto.ca/~datamining/dmc/logistic_regression.htm, last access 22.08.2017.
[19] Bicego, M. and Loog, M, “Weighted K-Nearest Neighbor revisited”, 23rd International Conference on
Pattern Recognition (ICPR), 2016, pp. 1642-1647.
[20] Ting, K. and Zheng, Z. “Improving the Performance of Boosting for Naive Bayesian Classification.
Methodologies for Knowledge Discovery and Data Mining”, 1999, pp.296-298.
[21] Peng Ye, “The decision tree classification and its application research in personnel management”,
Proceedings of 2011 International Conference on Electronics and Optoelectronics, 2011, pp. 1-4.
[22] Entropy of a decision tree classification algorithm [Online],
http://chemeng.utoronto.ca/~datamining/dmc/decision_tree.htm, last access 22.08.2017.
[23] Muda, Z., Yassin, W., Sulaiman, M. and Udzir, N. “Intrusion detection based on k-means clustering and
OneR classification”, 2011 7th International Conference on Information Assurance and Security (IAS).
[24] Kerdegari, H., Samsudin, K., Ramli, A. and Mokaram, S. “Evaluation of fall detection classification
approaches”, 2012 4th International Conference on Intelligent and Advanced Systems (ICIAS2012)
[25] Bigdeli, Y.; Barbato, M. Use of a low-cost concrete-like fluorogypsum-based blend for applications in
underwater and coastal protection structures. In Proceedings of the OCEANS 2017—Anchorage Conference,
Anchorage, AK, USA, 18–21 September 2017; pp. 1–5. Available online:
https://ieeexplore.ieee.org/abstract/document/8232181 (accessed on 5 May 2021).
[26] Reiter, L.; Wangler, T.; Anton, A.; Flatt, R.J. Setting on demand for digital concrete—Principles,
measurements, chemistry, validation. Cem. Concr. Res. 2020, 132, 106047. [CrossRef]
[27] Amran, Y.M.; Alyousef, R.; Alabduljabbar, H.; El-Zeadani, M. Clean production and properties of
geopolymer concrete; A review. J. Clean. Prod. 2020, 251, 119679. [CrossRef]

38 | Page
Publisher: Noida Institute of Engineering & Technology,
19, Knowledge Park-II, Institutional Area, Greater Noida (UP), India.