A review on machine learning

Shaveta *

Assistant Professor, DCSA, Guru Nanak College, Ferozepur Cantt, Punjab, India.

International Journal of Science and Research Archive, 2023, 09(01), 281–285

Publication history: Received on 18 April 2023; revised on 26 May 2023; accepted on 28 May 2023

Article DOI: https://doi.org/10.30574/ijsra.2023.9.1.0410

Abstract
Machine learning is a particular branch of artificial intelligence that teaches a machine how to learn, whereas artificial
intelligence (AI) is the general science that aims to emulate human abilities. An AI method called machine learning
teaches computers to learn from their past experiences. Machine learning algorithms don't rely on a predetermined
equation as a model, but instead "learn" information directly from data using computational techniques. As the quantity
of learning examples increases, the algorithms adaptively get better at what they do. This paper provides an overview
of the field as well as a variety of machine learning approaches, including supervised, unsupervised, and reinforcement
learning and various languages used for machine learning.

Keywords: Machine learning; learning; Supervised learning; Artificial intelligence; Languages

1. Introduction
A developing technology called machine learning makes it possible for computers to learn autonomously from historical
data. Machine learning uses a variety of techniques to create mathematical models and make predictions based on
previous information or data. Currently, it is utilised for many different things, including recommender systems, email
filtering, Facebook auto-tagging, image identification, and speech recognition [1].

Modern machine learning differs from machine learning in the past due to new computing technologies. Researchers
interested in artificial intelligence wanted to discover if computers could learn from data; pattern recognition and the
idea that they can learn without being programmed to do certain tasks gave rise to it. Because models may
independently adjust as they are exposed to fresh data, machine learning's iterative component is crucial. In order to
provide trustworthy, reproducible decisions and results, they learn from earlier calculations. It's an old science, but it
has recently garnered new momentum [2]. A subset of artificial intelligence known as "machine learning" is stated to
focus on creating algorithms that enable computers to independently learn from data and past experiences. In 1959,
Arthur Samuel coined the phrase "machine learning."

Machine learning algorithms create a mathematical model with the aid of historical sample data, or "training data," that
aids in making predictions or judgements without being explicitly programmed. Computer science and statistics are
used with machine learning to create prediction models. Algorithms that learn from past data are created by machine
learning or used in it [3]. The performance will be higher the more information we supply.

Whenever new data is received, a machine learning system predicts the outcome for it by learning from historical data,
creating prediction models, and learning from past data. The volume of data used to develop the model influences how
well it predicts the output, as a larger volume of data allows for more accurate model construction.


Corresponding author: Shaveta
Copyright © 2023 Author(s) retain the copyright of this article. This article is published under the terms of the Creative Commons Attribution Liscense 4.0.
 International Journal of Science and Research Archive, 2023, 09(01), 281–285

Instead of coding code specifically for a difficult scenario where we need to make predictions, let's say we merely need
to give the data to generic algorithms, which then use the data to build logic and anticipate the results. Machine learning
has altered the way we approach the issue [4].

Machine learning is becoming increasingly necessary as time goes on. Because machine learning can perform activities
that are too complex for a person to carry out directly, it is necessary. Since we cannot access the vast quantity of data
manually as humans, we need computer systems, and here is where machine learning comes in to simplify our lives.

By giving machine learning algorithms a vast amount of data, we can train them to examine the data, build models, and
anticipate the desired output automatically. The cost function can be used to estimate how well the machine learning
algorithm performs given the amount of data. We can save time and money by using machine learning. Its use cases
make it simple to understand the significance of machine learning. Machine learning is currently employed in self-
driving cars, cyber fraud detection, face recognition, and Facebook friend suggestion, among other applications [5].
Several leading businesses, including Netflix and Amazon, have developed machine learning algorithms that make use
of a tremendous quantity of data to analyse customer interest and make product recommendations.

Because it can solve issues at a speed and scale that cannot be matched by the human mind alone, ML has been shown
to be useful. Machines can be trained to recognise patterns in and relationships between incoming data by putting large
amounts of processing power behind a single activity or a number of focused tasks. This allows machines to automate
repetitive tasks.

Machine learning can be broadly divided into three categories:

 Supervised learning
 Unsupervised learning
 Reinforcement in learning

2. Supervised learning
One form of machine learning technique is supervised learning, in which we train the machine learning system using
sample labelled data and then watch as it predicts the outcome.

In order to anticipate future events, supervised machine learning algorithms use labelled examples to apply what they
have learned in the past to fresh data. The learning method creates an inferred function to forecast output values by
examining a known training dataset [6]. After adequate training, the system may provide targets for any new input. In
order to identify flaws and correct the model as necessary, it can also compare its output with that which is proper and
intended.

In order to comprehend the datasets and learn about each one, the system builds a model using labelled data. After
training and processing, the model is tested by being given a sample set of data to see if it predicts the output accurately
or not [7].

The mapping of input and output data is the aim of supervised learning. A pupil learning under the teacher's supervision
is the same as supervised learning because it is based on supervision. Spam filtering serves as an illustration of
supervised learning. Supervised learning can be grouped further in two categories of algorithms:

2.1. Classification
When the output variable is categorical, meaning there are two classes, such as Yes-No, Male-Female, True-False, spam
detection, etc., classification methods are used. Several Classification algorithms are: Random Forest, Decision Trees,
Logistic Regression, Support Vector Machines

2.2. Regression
Regression procedures are applied if there is a correlation between the input and output variables. It is used to forecast
continuous variables like weather, market trends, and other things. Several well-liked regression algorithms that fall
under supervised learning are listed below: Linear Regression, Regression Trees, Non-Linear Regression, Bayesian
Linear Regression, Polynomial Regression

282
 International Journal of Science and Research Archive, 2023, 09(01), 281–285

3. Unsupervised learning
Unsupervised learning is a type of learning where a computer picks up information without any human intervention.
The machine is trained using a set of unlabeled, unclassified, or uncategorized data, and the algorithm is required to
respond independently to that data [8]. Unsupervised learning's objective is to reorganize the incoming data into fresh
features or a collection of objects with related patterns. There is no predefined outcome in unsupervised learning.

When training data is neither categorised nor labelled, unsupervised machine learning techniques are utilised.
Unsupervised learning investigates how systems might extrapolate a function from unlabeled data to describe a hidden
structure. The system can never be guaranteed that the output is correct. Instead, it infers what the result should be
based on datasets. The machine searches through the vast volume of data for helpful insights. It can also be divided into
two types of algorithms:

3.1. Clustering
Using the clustering technique, items are grouped into clusters so that those who share the most similarities stay in one
group and share little to none with those in another. The data objects are classified based on the existence or lack of
commonalities discovered by cluster analysis.

3.2. Association
An unsupervised learning technique called an association rule is used to uncover the connections among the variables
in a sizable database. It establishes the group of items that co-occur in the collection. Marketing strategy is more effective
because to the association rule. People who buy X (let's say, bread) also frequently buy Y (let's say, butter or jam). Market
Basket Analysis is an illustration of an association rule in action.

4. Reinforcement learning
A learning agent in a reinforcement learning system receives a reward for each correct action and receives a penalty for
each incorrect activity. With the help of this feedback, the agent automatically learns and performs better [9]. The agent
explores and engages with the environment during reinforcement learning. An agent performs better since its objective
is to accrue the most reward points.

Algorithms for reinforcement learning interact with their surroundings by taking actions and identifying successes or
failures. Trial-and-error learning and delayed rewards are two of reinforcement learning's most important features.
With the help of this technique, machines and software agents may automatically select the best course of action in a
given situation to enhance performance. The reinforcement signal, which is straightforward reward feedback, is
necessary for the agent to figure out which action is better [10]. Reinforcement learning is demonstrated by the robotic
dog, which automatically learns how to move its arms [11].

5. Languages used for Machine Learning

5.1. Python
A fantastic programming language is Python. Although it is easy to read, it is also capable of a wide range of tasks. Unlike
other languages, it enables quick iteration. If you only need to alter one or two lines of your code, you can immediately
run the changed version. For minor modifications, there's no need to update numerous locations. No need to wait
minutes or hours for your code to be recompiled before running it. Python has one of the most straightforward syntaxes
and one of the most natural languages. Learning, reading, and correcting errors are therefore simple. Python is an open-
source language, which is the best part [12]. Hence, both accessing and distributing it are free. It consequently features
flexible libraries. The burgeoning fields of data science, AI, and machine learning can benefit from this language. Python
is anticipated to continue to be in high demand in the future because of this use. It is often used in web-based
applications as well. Additionally, you can use it to create desktop and mobile applications [13].

A library is a group of programs or modules created to carry out particular tasks. Among other things, some modules
can assist with the graphical user interface, data analysis, and graphics. You can import the code from the library rather
than having to write it each time you need it.

283