INTERNSHIP REPORT

ON
MATLAB WITH MACHINE LEARNING

A report submitted in partial fulfilment of the requirements for the award

of degree of
Bachelor Of Technology(B.TECH)

DEPARTMENT OF POWER ELECTRONICS AND INSTRUMENTATION

JORHAT INSTITUTE OF SCIENCE AND TECHNOLOGY
JORHAT,ASSAM
AFFILIATED TO ASSAM SCIENCE AND TECHNOLOGY UNIVERSITY

Submitted by-
Priya Saikia(210810017026)
Pranamika Bora(220810017030)
Issa Gowala(220810017017)

1
 CERTIFICATE OF APPROVAL

The undersigned certify that Priya Saikia, Pranamika Bora and Isha Gowala of 5th semester
from the Department of Power Electronics And Instrumentation Engineering have done the
internship under my guidance during the period from 18-06-2024 to 17-08-23.

Dr. Nilakshi Devi

Guide and Faculty,
IEEE, Guwahati University, Guwahati

Mrs. Chandana Dev

Head Of Department ,
Power Electronics And Instrumentation
Jorhat Institute Of Science And Technology

2
 DECLARATION

I hereby declare that the internship report entitled “Matlab With Machine Learning” submitted
for the B.TECH degree in Department of Power Electronics And Instrumentation is my original
work and has not been formed the basis for the award of any other degree, diploma, fellowship
or any other similar titles.

Signature of student
Place:
Date:

3
 ACKNOWLEDGEMENT

It gives us immense pleasure to express our deepest sense of gratitude and sincere thanks to
our highly respected and esteemed guide Dr. Nilakshi Devi. Her useful suggestions are
sincerely acknowledged.
We would also like to express our sincere thanks to Mr. Angshuman Goswami(TPO) for giving
us this opportunity to undertake this internship and for his wholehearted support.

We are also grate to our respected HOD and teachers for their constant support and guidance.
At the end we would like to express our sincere thanks to all our friends and other who helped
us directly or indirectly during this internship work.

4
 ABSTRACT

This report explores the seamless integration of machine learning algorithms within MATLAB,
highlighting its capabilities, applications, and performance. MATLAB’s machine learning
toolbox provides a comprehensive platform for data preprocessing, feature selection, model
development, and evaluation. The report delves into the implementation of supervised and
unsupervised learning techniques, including neural networks, decision trees, and clustering.
Real-world case studies demonstrate the efficacy of MATLAB’s machine learning tools in
image classification, predictive analytics, and signal processing. The results showcase the
efficiency and accuracy of MATLAB’s machine learning algorithms, underscoring its potential
in various domains such as engineering, computer vision, and data science. This report serves
as a valuable resource for researchers, engineers, and students seeking to leverage MATLAB’s
machine learning capabilities.

5
 TABLE OF CONTENTS
CONTENTS PAGE NO

1. Institute Description
2. Methodology
3. Introduction to Machine Learning
4. Getting Started with Matlab
5. Understanding the basics of Machine Learning
6. Matlab Toolboxes for Machine Learning
7. Data Prepocessing in MATLAB
8. Supervised Learning Algorithms
9. Unsupervised Learning Algorithms
10. Evaluating Machine Learning Models in Matlab
11. Deep Learning With Matlab
12. Advanced Topics in Machine Learning with MATLAB
13. Real-World Applications of Machine Learning in MATLAB
14.MATLAB for Model Deployment and Automation
15. Challenges and Future Directions
16. Conclusion
17. References

6
 INSTITUTE DESCRIPTION

Guwahati University, officially known as The University of Guwahati, is a

premier public research university located in Guwahati, Assam, India.
Established in 1948, it is one of the oldest and most prestigious institutions of
higher education in the northeastern region of India. The university is known for
its academic excellence and diverse range of programs across various
disciplines, including arts, sciences, engineering, social sciences, business, law,
and more.

7
 METHODOLOGY

This report discusses mainly about the study of MACHINE LEARNING

WITH MATLAB. This study was conducted with the help of faculty
members of Guwahati University,Guwahati with the help of modern
computers. This programme provided us prolonged and tailored help to
understand and learn the use of Machine Learning with MATLAB in the
real world. We allocated ourselves in a well-recognised computer institute
for a crash course with the duration period of time.

We were taught from the rock bottom to a notably well state. This course
has also taught us the importance of Machine Learning with MATLAB in
an individuals life in implementing his ideas.

After completion of our course, we were provided with valuable

certificates.

8
 INTRODUCTION

What is Machine Learning?

Machine learning (ML) is a type of Artificial Intelligence (AI) that allows

computers to learn without being explicitly programmed. It involves
feeding data into algorithms that can then identify patterns and make
predictions on new data. Machine learning is used in a wide variety of
applications, including image and speech recognition, natural language
processing, and recommender systems.

Definition of Learning

A computer program is said to learn from experience E concerning some

class of tasks T and performance measure P, if its performance at tasks T,
as measured by P, improves with experience E.
Examples
 Handwriting recognition learning problem
o Task T : Recognizing and classifying handwritten words within
images
o Performance P : Percent of words correctly classified
o Training experience E : A dataset of handwritten words with
given classifications
 A robot driving learning problem
o Task T : Driving on highways using vision sensors
o Performance P : Average distance traveled before an error
o Training experience E : A sequence of images and steering
commands recorded while observing a human driver
o Classification of Machine Learning
o Machine learning implementations are classified into four major
categories, depending on the nature of the learning “signal” or
“response” available to a learning system which are as follows:
o 1. Supervised learning:
o Supervised learning is the machine learning task of learning a
function that maps an input to an output based on example input-
output pairs. The given data is labeled.
Both classification and regression problems are supervised
learning problems.

9
 o Example – Consider the following data regarding patients entering
a clinic . The data consists of the gender and age of the patients and
each patient is labeled as “healthy” or “sick”.

o Gender o Age o Label

o M o 48 o sick

o M o 67 o sick

o F o 53 o healthy

o M o 49 o sick

o F o 32 o healthy

o M o 34 o healthy

o M o 21 o healthy

2. Unsupervised learning:
Unsupervised learning is a type of machine learning algorithm used to draw
inferences from datasets consisting of input data without labeled responses. In
unsupervised learning algorithms, classification or categorization is not included
in the observations. Example: Consider the following data regarding patients
entering a clinic. The data consists of the gender and age of the patients.
Gender Age

M 48

M 67

F 53

10
 M 49

F 34

M 21

As a kind of learning, it resembles the methods humans use to figure out that certain
objects or events are from the same class, such as by observing the degree of
similarity between objects. Some recommendation systems that you find on the
web in the form of marketing automation are based on this type of learning.

3. Reinforcement Learning:

Reinforcement learning is the problem of getting an agent to act in the world so

as to maximize its rewards.

A learner is not told what actions to take as in most forms of machine learning
but instead must discover which actions yield the most reward by trying them.
For example — Consider teaching a dog a new trick: we cannot tell him what to
do, what not to do, but we can reward/punish it if it does the right/wrong thing.

When watching the video, notice how the program is initially clumsy and
unskilled but steadily improves with training until it becomes a champion.

4. Semisupervised Learning:

Where an incomplete training signal is given: a training set with some (often
many) of the target outputs missing. There is a special case of this principle
known as Transduction where the entire set of problem instances is known at
learning time, except that part of the targets are missing. Semi-supervised
learning is an approach to machine learning that combines small labeled data with
a large amount of unlabeled data during training. Semi-supervised learning falls
between unsupervised learning and supervised learning.

11
ff

12
 FIG: Machine Learning Lifecycle

4. GETTING STARTED WITH MATLAB

13
 Introduction

MATLAB (Matrix Laboratory) is a high-level programming language and

environment primarily designed for numerical computation, data analysis,
algorithm development, and visualization. It provides a rich set of tools for
mathematical computations, and is widely used in engineering, science, finance,
and academic research. This report provides an overview of the key features of
MATLAB and guides users on how to get started with basic operations,
scripting, and plotting.

1. Installation of MATLAB
To begin using MATLAB, it is essential to install the software on your
computer. MATLAB is available as a paid version, but there are free trials and
academic licenses available.

Steps to Install MATLAB:

1. Visit the official MathWorks

website at www.mathworks.com.
2. Create a MathWorks account if you do not already have one.

3. Download and install the latest version of MATLAB.

4. Follow the on-screen instructions to complete the installation process.

14
 2. MATLAB Interface Overview

Upon launching MATLAB, users are greeted with the main interface,
which consists of several key components:

2.1 Command Window

The Command Window is the primary interface for typing and executing
commands. It allows users to interact directly with MATLAB, execute
one-line commands, and display results.

2.2 Workspace

The Workspace area displays all the variables that have been defined
during the current session. This allows users to view and manage
variables and their values.

2.3 Current Folder

The Current Folder window shows the files and directories in the current
working directory. It enables users to navigate and manage their
MATLAB files.

2.4 Editor

The Editor is used to create, edit, and run scripts and functions. It allows
users to write more complex programs and organize their work into
reusable code blocks.

3. Basic MATLAB Operations

MATLAB's primary strength lies in its ability to perform complex

numerical calculations with ease. Below are examples of basic operations
that can be performed in MATLAB.

3.1 Arithmetic Operations

MATLAB supports basic arithmetic operations, such as addition,

subtraction, multiplication, and division:

15
 a = 5 + 3; % Addition
b = 10 - 2; % Subtraction
c = 4 * 6; % Multiplication
d = 8 / 2; % Division

3.2 Variables

MATLAB allows users to define variables to store values for later use:

x = 10; % Assigning value 10 to variable x

y = 20; % Assigning value 20 to variable y
z = x + y; % Adding x and y, and storing result in z

3.3 Matrices and Arrays

MATLAB is particularly powerful for matrix and array operations. Users

can define matrices and perform various operations, such as addition,
multiplication, and element-wise operations:

A = [1 2; 3 4]; % 2x2 matrix

B = [5 6; 7 8]; % Another 2x2 matrix

C = A + B; % Matrix addition
D = A * B; % Matrix multiplication
E = A .* B; % Element-wise multiplication

4. Plotting and Visualization

MATLAB provides extensive capabilities for visualizing data through

various plotting functions. A basic plot can be created with the plot
function, as shown in the example below:

Example: Plotting Sine and Cosine Functions

x = 0:0.1:2*pi; % Create a vector from 0 to 2*pi with steps of 0.1

y1 = sin(x); % Compute the sine of x
y2 = cos(x); % Compute the cosine of x

plot(x, y1, '-r', x, y2, '-b'); % Plot sine in red, cosine in blue

16
 legend('sin(x)', 'cos(x)');
title('Sine and Cosine Functions');
xlabel('x');
ylabel('y');
grid on;

This code generates a plot of the sine and cosine functions with labeled
axes, a title, and a legend.

5. Writing Scripts and Functions

5.1 Scripts

A script in MATLAB is a file containing a series of commands that are

executed sequentially. Scripts are useful for automating tasks and
performing repetitive operations.

To create a script:

1. Open the MATLAB Editor by clicking "New Script."

2. Write your MATLAB code in the editor.
3. Save the file with a .m extension (e.g., my_script.m).
4. Run the script by typing its name in the Command Window.

5.2 Functions

Functions are similar to scripts but allow for input and output arguments.
They enable users to write reusable blocks of code.

A basic MATLAB function might look like this:

function output = squareNumber(input)

output = input^2; % Returns the square of the input
end

This function can be saved in a file named squareNumber.m and called

from the Command Window or other scripts to compute the square of a
number.

17
 6. MATLAB Documentation and Help System

MATLAB provides an extensive documentation system that users can

consult for assistance with functions, syntax, and examples. Users can
access help directly within the environment using the following
commands:

Help on a specific function: help function_name

Example: help plot

Full documentation: doc function_name

Example: doc plot

FIG: DOC PLOT OF MATLAB SIMULATION CURVE

UNDERSTANDING THE BASICS OF MACHINE LEARNING

1. Key Concepts in Machine Learning

18
 Machine Learning involves several foundational concepts, each
contributing to the ability of a system to learn and generalize from
data.

1.1. Data

Data is the cornerstone of machine learning. The more high-

quality data a model is exposed to, the better it can learn. Data
typically comes in two forms:

Features (Input Variables): The attributes or characteristics that

are used to predict an outcome (e.g., age, income, weather
conditions).

Labels (Output Variables): The target variable or outcome that we

aim to predict or classify (e.g., whether a customer will buy a
product, the price of a house).

1.2. Model

A model in ML is a mathematical representation of a real-world

process. It maps input features to output labels. The goal of
training a model is to optimize its parameters to minimize the
difference between predicted and actual values (known as the
error or loss).

1.3. Training and Testing Data

19
 To assess the performance of an ML model, the data is often split
into two sets:

Training Data: Used to train the model and adjust its parameters.

Testing Data: Used to evaluate the model's ability to generalize to

unseen data, helping to avoid overfitting (where the model
becomes too tailored to the training data).

1.4. Learning Algorithms

Learning algorithms are used to adjust the model’s parameters by

minimizing the error between the predicted and actual outcomes.
Common algorithms include linear regression, decision trees, and
neural networks.

2. Types of Machine Learning

Machine learning is generally classified into three primary types

based on the learning task and the structure of the data.

2.1. Supervised Learning

Supervised learning is the most common form of machine

learning, where the model is trained on labeled data. In this
approach, the algorithm learns a mapping from input features (X)
to output labels (Y). The goal is to make accurate predictions for
new, unseen data.

Common Algorithms:

20
 Linear Regression: Used for predicting continuous values (e.g.,
predicting house prices).

Classification: Used for predicting categorical outcomes (e.g.,

classifying emails as spam or not spam). Popular algorithms
include Decision Trees, Support Vector Machines (SVM), and k-
Nearest Neighbors (k-NN).

2.2. Unsupervised Learning

In unsupervised learning, the model is trained on data that has no

labeled outcomes. The goal is to uncover hidden patterns or
structures within the data. Common applications include
clustering and dimensionality reduction.

Common Algorithms:

K-Means Clustering: Groups data points into clusters based on

their similarity.

Principal Component Analysis (PCA): Reduces the

dimensionality of large datasets while preserving as much
variance as possible.

2.3. Reinforcement Learning

Reinforcement learning (RL) is an area where an agent learns to

make decisions by interacting with an environment. The agent
receives feedback through rewards or penalties, which guide it to
improve its performance over time. RL is often used in robotics,
game-playing AI (e.g., AlphaGo), and autonomous systems.

2.4. Semi-supervised and Self-supervised Learning

21
 These are hybrid approaches where the model is trained on a
small amount of labeled data and a large amount of unlabeled
data. Semi-supervised learning improves learning efficiency by
leveraging both types of data.

6. MATLAB TOOLBOXES FOR MACHINE LEARNING

22
Statistics and Machine Learning Toolbox
The Statistics and Machine Learning Toolbox is perhaps the most
comprehensive toolbox for general-purpose machine learning tasks in
MATLAB. It provides a wide range of algorithms for supervised and
unsupervised learning, as well as tools for model evaluation and
assessment.
Key Features:
Supervised Learning: Functions for classification, regression, and
support vector machines (SVM). Models such as decision trees, k-
nearest neighbors (k-NN), and logistic regression are supported.
Unsupervised Learning: Clustering algorithms like k-means,
hierarchical clustering, and Gaussian mixture models (GMM) are
included.
Dimensionality Reduction: Tools like principal component analysis
(PCA) and t-distributed stochastic neighbor embedding (t-SNE) help
reduce the number of input features while preserving important
patterns.
Model Evaluation: Provides cross-validation, confusion matrices, and
performance metrics (accuracy, precision, recall, F1 score) for model
evaluation.
Example Use Case:

Predicting Customer Churn: You can use classification algorithms such

as decision trees or SVM to predict whether a customer will churn,
based on historical data.

2.2. Deep Learning Toolbox

23
The Deep Learning Toolbox is designed for developing deep neural
networks (DNNs), including convolutional neural networks (CNNs),
recurrent neural networks (RNNs), and other advanced neural
architectures. This toolbox is ideal for image recognition, speech
processing, and time-series analysis.
Key Features:
Network Architectures: Tools for defining, training, and validating
deep neural networks, including CNNs, RNNs, and autoencoders.
Pretrained Models: A library of pretrained models for tasks like image
classification, object detection, and natural language processing, which
can be fine-tuned to specific datasets.
GPU Acceleration: The toolbox supports GPU-based training for
faster model development, particularly useful when working with large
datasets.

Transfer Learning: Allows the reuse of a pretrained model on a new

task, which can significantly reduce training time and improve model
performance on small datasets.

Example Use Case:

Image Classification: Using CNNs to classify images of animals or
medical scans. You can fine-tune a pretrained ResNet model for
specific image datasets.

2.3. Computer Vision Toolbox

24
The Computer Vision Toolbox extends MATLAB's capabilities for
machine learning tasks specifically related to computer vision, such as
object detection, image segmentation, and feature extraction.
Key Features:
Object Detection: Implements algorithms for detecting objects in
images and video streams, such as Haar-like features and deep learning-
based approaches.
Image Processing: Includes functions for image filtering, edge
detection, and transformation, which are often required in computer
vision pipelines before applying machine learning models.
Feature Extraction: Tools for extracting key features from images,
such as SIFT, SURF, and HOG (Histogram of Oriented Gradients),
which can then be used as inputs to machine learning models.
Video Analysis: Functions for analyzing video data, detecting moving
objects, and tracking them across frames.
Example Use Case:
Face Recognition: Detecting faces in real-time from video feeds and
classifying them using a machine learning model.

2.4. Reinforcement Learning Toolbox

The Reinforcement Learning Toolbox is designed for developing and

training reinforcement learning (RL) models. In RL, an agent learns to
make decisions by interacting with its environment and receiving
rewards or penalties.

Key Features:
25
Q-Learning and Deep Q-Networks (DQN): Supports classic RL
algorithms like Q-learning as well as advanced techniques such as deep
Q-networks (DQN) for more complex tasks.
Simulations: Allows for integration with simulation environments
(e.g., Simulink, Gazebo) to train RL agents in a controlled environment.
Policy Optimization: Includes algorithms for policy optimization,
such as proximal policy optimization (PPO) and trust region policy
optimization (TRPO).
Example Use Case:
Autonomous Driving: Using reinforcement learning to train an
autonomous vehicle agent to navigate through traffic and avoid
obstacles.

2.5. Natural Language Processing Toolbox

The Natural Language Processing Toolbox provides specialized
functions for processing and analyzing text data, which is crucial for
tasks like sentiment analysis, text classification, and language
modeling.
Key Features:
Text Preprocessing: Tools for tokenizing, stemming, and removing
stop words from text data.
Topic Modeling: Algorithms for extracting topics from large corpora
of text, such as Latent Dirichlet Allocation (LDA).
Text Classification: Methods for training models to classify text data,
such as Naive Bayes and deep learning-based approaches.

7. DATA PREPOCESSING IN MATLAB

26
MATLAB offers various built-in functions and toolboxes for
addressing different aspects of data preprocessing. Below are the most
common tasks involved in data preparation.

2.1. Handling Missing Data

Missing data is a common issue in real-world datasets and can
significantly affect the performance of machine learning models.
MATLAB provides several methods for detecting and handling missing
values.
Key Functions:
Detect Missing Values: The isnan() function identifies missing values
in numeric datasets.
missing_values = isnan(data);
Fill Missing Values: The fillmissing() function can replace missing
values with a specified value, such as the mean, median, or
interpolation.
data_filled = fillmissing(data, 'constant', 0); % Replace with zero
data_filled = fillmissing(data, 'linear'); % Linear interpolation
Example Use Case:
Time-Series Data: In time-series analysis, missing values can be filled
using linear interpolation to maintain continuity in the data.

2.2. Data Normalization and Scaling

Data normalization and scaling ensure that numerical features are on a
similar scale, which helps machine learning algorithms perform better,
particularly distance-based algorithms like k-nearest neighbors (k-NN)
or gradient descent-based models.

27
Key Functions:
Normalization: The normalize() function scales features to a specific
range (typically [0, 1]).
normalized_data = normalize(data);
Standardization: To standardize features (zero mean and unit variance),
you can use zscore().
standardized_data = zscore(data);
Example Use Case:
Customer Data: If you have features like age (range 0-100) and
income (range 10,000-100,000), normalization ensures that both
features contribute equally when input into a machine learning model.

2.3. Encoding Categorical Variables

Many machine learning algorithms require numerical input, so
categorical data (e.g., gender, country, product type) must be converted
into numeric values. MATLAB provides several methods for encoding
categorical variables.
Key Functions:
Categorical to Numeric Conversion: The grp2idx() function maps
categorical data to integer indices.
categories = {'male', 'female', 'female', 'male'};
numeric_data = grp2idx(categories);
One-Hot Encoding: MATLAB’s dummyvar() function creates binary
vectors (one-hot encoding) for categorical variables.
categorical_data = categorical({'apple', 'banana', 'apple', 'orange'});
one_hot_data = dummyvar(categorical_data);
Example Use Case:

28
Survey Data: If you have survey responses with categories like “Yes”
or “No,” you can convert these to numeric values (e.g., 1 and 0) for use
in machine learning models.
2.4. Feature Extraction and Selection
Feature extraction involves creating new features from existing data to
improve the predictive power of a model. Feature selection, on the
other hand, involves choosing the most relevant features from a larger
set, reducing dimensionality and preventing overfitting.
Key Functions:
Principal Component Analysis (PCA): MATLAB’s pca() function
performs dimensionality reduction by projecting data onto a set of
orthogonal components that capture the maximum variance in the data.
[coeff, score, latent] = pca(data);
Correlation-Based Feature Selection: You can identify highly
correlated features using corr() to ensure that redundant features are
removed.
correlation_matrix = corr(data);
Example Use Case:
Medical Data: When working with datasets containing thousands of
features (such as genetic data), PCA can help reduce the dimensionality
and retain the most important information.

8. SUPERVISED LEARNING ALGORITHMS

29
Supervised learning involves training a model on labeled data to predict
outcomes for unseen data. MATLAB provides a comprehensive set of
tools to implement a variety of supervised learning algorithms, ranging
from simple models like linear regression to more complex ones such
as neural networks and support vector machines (SVM). This report
explores key supervised learning algorithms in MATLAB and their
applications.
Common Supervised Learning Algorithms in MATLAB
1. Linear Regression
Linear regression is used for predicting continuous values based on
input features. In MATLAB, the fitlm() function fits a linear regression
model.

mdl = fitlm(X, Y); % X: input features, Y: target variable

Use Case: Predicting house prices based on size, location, etc.

2. Logistic Regression
Logistic regression is used for binary classification. MATLAB uses the
mnrfit() function for this task.

mdl = mnrfit(X, Y, 'model', 'binary'); % X: input data, Y: binary target

Use Case: Classifying emails as spam or not.

3. Decision Trees

30
Decision trees partition data into subsets based on feature values. In
MATLAB, use fitctree() for classification and fitrtree() for regression.

tree = fitctree(X, Y); % X: features, Y: class labels

Use Case: Medical diagnosis based on patient data.

4. Support Vector Machines (SVM)

SVM is a powerful classification algorithm that finds the optimal

hyperplane separating classes. Use fitcsvm() in MATLAB.

svm = fitcsvm(X, Y); % X: input data, Y: target labels

Use Case: Image classification tasks.

5. k-Nearest Neighbors (k-NN)

k-NN classifies data based on the majority class of its nearest
neighbors. MATLAB’s fitcknn() is used to train the model.
knn = fitcknn(X, Y, 'NumNeighbors', 5);
Use Case: Customer segmentation based on purchasing behavior.

9.UNSUPERVISED LEARNING ALGORITHMS

31
1. 1. K-Means Clustering

K-means is one of the most popular clustering algorithms, which

partitions data into k clusters based on feature similarity. Each data
point is assigned to the nearest cluster center.

Key Function: kmeans()

[idx, C] = kmeans(X, k); % X: data, k: number of clusters

Use Case: Customer segmentation based on purchasing behavior.

2. Hierarchical Clustering

Hierarchical clustering builds a tree-like structure (dendrogram) to

represent data. It can be used for both agglomerative and divisive
clustering approaches.

Key Function: linkage() and dendrogram()

Z = linkage(X, 'ward'); % X: data, 'ward' for minimizing variance

dendrogram(Z);

Use Case: Grouping similar gene expressions in biological data.

32
3. Gaussian Mixture Models (GMM)

Gaussian Mixture Models assume that the data is generated from a

mixture of several Gaussian distributions. GMM is useful for
modeling data that has multiple sub-populations.

Key Function: fitgmdist()

GMModel = fitgmdist(X, k); % X: data, k: number of components

Use Case: Image segmentation where each region is modeled as a

Gaussian distribution.

4. Principal Component Analysis (PCA)

PCA is a dimensionality reduction technique that projects data into a

lower-dimensional space while retaining most of the variance in the
data. PCA is often used for feature extraction or to visualize high-
dimensional data.

Key Function: pca()

[coeff, score, latent] = pca(X); % X: data

Use Case: Reducing the dimensions of financial data while preserving

essential information.

33
5. t-Distributed Stochastic Neighbor Embedding (t-SNE)

t-SNE is a non-linear dimensionality reduction technique primarily

used for visualizing high-dimensional data in 2D or 3D space by
minimizing divergence between probability distributions.

Key Function: tsne()

Y = tsne(X); % X: high-dimensional data

Use Case: Visualizing clusters of handwritten digits in a lower-

dimensional space.

EVALUATING MACHINE LEARNING MODELS IN

MATLAB

34
 Evaluating the performance of machine learning models is a
crucial step in ensuring that the model generalizes well to new,
unseen data. In MATLAB, several techniques and metrics are
available for assessing model performance, including accuracy,
precision, recall, F1-score, and cross-validation. This report
highlights key methods and functions in MATLAB for
evaluating machine learning models, particularly focusing on
classification and regression tasks.

1. Model Evaluation for Classification

1.1. Confusion Matrix

A confusion matrix provides a detailed breakdown of

classification model performance, showing the counts of true
positives, true negatives, false positives, and false negatives. It
helps in understanding how well the model distinguishes
between classes.

Key Function: confusionmat()

cm = confusionmat(Y_true, Y_pred); % Y_true: actual labels,

Y_pred: predicted labels

Metrics derived:

Accuracy: Proportion of correct predictions.

accuracy = sum(diag(cm)) / sum(cm(:));

Precision, Recall, and F1-Score: These metrics give more

insight, particularly when dealing with imbalanced classes. You
can compute them manually or use classificationReport() from
the Statistics and Machine Learning Toolbox.
35
 1.2. Receiver Operating Characteristic (ROC) Curve and
AUC

The ROC curve plots the true positive rate (recall) against the
false positive rate (1-specificity). The area under the ROC curve
(AUC) is a performance measure, where higher values (close to
1) indicate better models.

Key Function: rocmetrics() and plotroc()

[X, Y, T, AUC] = perfcurve(Y_true, scores, 'trueclass', 1); %

scores: probability scores from model
plot(X, Y); % Plot ROC curve

1.3. Cross-Validation

Cross-validation helps assess model performance by training

and testing on multiple data splits, reducing the chance of
overfitting. In k-fold cross-validation, the data is split into k
subsets, and the model is trained k times, each time with a
different test fold.

Key Function: crossval()

cv = cvpartition(length(Y), 'KFold', 5); % 5-fold cross-

validation
model = fitcsvm(X, Y); % Example for SVM
cvmodel = crossval(model, 'CVPartition', cv);

Cross-validation returns the model's performance across

different subsets of the data.
DEEP LEARNING WITH MATLAB

36
 Deep learning is a subset of machine learning focused on algorithms
inspired by the structure and function of the brain, primarily artificial
neural networks. MATLAB provides a robust environment for building,
training, and deploying deep learning models with its Deep Learning
Toolbox. This toolbox offers prebuilt functions and tools for designing
deep neural networks, including feedforward networks, convolutional
networks, and recurrent networks. In this report, we explore key deep
learning workflows in MATLAB.

Key Components of Deep Learning in MATLAB

1. Creating Neural Networks

MATLAB allows users to build deep learning models using layers such as
fully connected, convolutional, and LSTM (long short-term memory)
layers.

Key Functions:

layer: Create individual layers.

LayerGraph: Combine layers to create a neural network architecture.

dlnetwork: Define a custom deep learning network.

Example: Simple Feedforward Neural Network

layers = [
fullyConnectedLayer(10)
reluLayer
fullyConnectedLayer(1)
regressionLayer
];

2. Training Neural Networks

MATLAB provides several functions to train deep learning models, using

datasets such as images, sequences, and text. The training process

37
 involves adjusting the weights of the network to minimize the loss
function.

Key Function: trainNetwork()

options = trainingOptions('sgdm', 'MaxEpochs', 20, 'InitialLearnRate',

0.01);
model = trainNetwork(XTrain, YTrain, layers, options);

Here, XTrain and YTrain are the input data and labels, and options
defines the training parameters such as learning rate and epochs.

3. Transfer Learning

Transfer learning enables the use of pre-trained models for new tasks,
saving time and computational resources. MATLAB offers pretrained
networks such as AlexNet, VGG16, and ResNet, which can be fine-tuned
on a new dataset.

Key Functions:

alexnet, vgg16, resnet50: Load pre-trained networks.

layerGraph: Modify the network for the new task.

Example: Fine-Tuning a Pretrained Network

net = alexnet;
layers = net.Layers;
layers(end-3:end) = []; % Remove last layers
newLayers = [
fullyConnectedLayer(10)
softmaxLayer
classificationLayer
];
layers = [layers; newLayers];
model = trainNetwork(XTrain, YTrain, layers, options);

4. Convolutional Neural Networks (CNNs)

38
 CNNs are widely used for image classification, object detection, and
segmentation. MATLAB simplifies building CNNs with specialized
layers like convolutional, pooling, and batch normalization layers.

Key Function: convolution2dLayer(), maxPooling2dLayer()

layers = [
imageInputLayer([28 28 1])
convolution2dLayer(3, 16)
reluLayer
maxPooling2dLayer(2, 'Stride', 2)
fullyConnectedLayer(10)
softmaxLayer
classificationLayer
];

Use Case: Image classification, facial recognition.

5. Recurrent Neural Networks (RNNs)

Recurrent Neural Networks (RNNs) and their advanced variants, such as

LSTMs and GRUs, are ideal for sequence-based data such as time-series,
speech, or text.

Key Function: lstmLayer()

layers = [
sequenceInputLayer(1)
lstmLayer(50)
fullyConnectedLayer(1)
regressionLayer
];

Use Case: Predicting stock prices, speech recognition.

39
ADVANCED TOPICS IN MACHINE LEARNING WITH
MATLAB

1. Deep Learning and Neural Networks

Deep learning is one of the most important areas in modern machine

learning. MATLAB’s Deep Learning Toolbox allows the creation,
training, and evaluation of deep neural networks (DNNs), convolutional
neural networks (CNNs), recurrent neural networks (RNNs), and more.

Key Techniques:

Convolutional Neural Networks (CNNs): Used for tasks such as image

classification and object detection, CNNs are built using convolutional
and pooling layers to capture spatial hierarchies in images.

Recurrent Neural Networks (RNNs): RNNs, including LSTMs (Long

Short-Term Memory networks) and GRUs (Gated Recurrent Units), are
particularly suited for sequential data such as time-series forecasting,
language modeling, and speech recognition.

Autoencoders: Unsupervised deep learning networks used for

dimensionality reduction, anomaly detection, and feature
extraction.

Tools:

trainNetwork(): To train deep learning models.

Pre-trained models: MATLAB provides access to networks like

AlexNet, ResNet, and VGG for transfer learning.

2. Transfer Learning
40
 Transfer learning is an advanced technique where a model pre-
trained on a large dataset is adapted for a new, often smaller,
dataset. This reduces training time and computational resources,
particularly when data is scarce.

Key Steps:

1. Load a Pre-trained Model: Load models like ResNet,

AlexNet, or VGG16.

2. Modify the Model: Replace the final layers to fit your new
task.

3. Fine-Tuning: Train the modified model on your dataset,

adjusting only certain layers to minimize overfitting.

net = alexnet; % Load pre-trained network

layers = net.Layers;
layers(end-3:end) = []; % Modify the output layer
new_layers = [fullyConnectedLayer(10), softmaxLayer,
classificationLayer];
layers = [layers; new_layers];
trainedModel = trainNetwork(XTrain, YTrain, layers, options);

REAL-WORLD APPLICATIONS OF MACHINE LEARNING

WITH MATLAB

41
1. Healthcare and Medical Diagnosis
Medical Image Analysis: MATLAB is widely used to analyze medical
images (such as MRI, CT scans, X-rays) for disease detection,
segmentation, and classification. ML algorithms help in tasks like
identifying tumors, detecting fractures, or classifying tissue types.

Predictive Healthcare: Machine learning models can be trained to

predict patient outcomes (e.g., disease progression, readmission risk)
using patient data such as demographics, lab results, and clinical
histories.

2. Autonomous Vehicles

Self-Driving Cars: MATLAB is used to develop machine learning

models that power autonomous driving systems, including object
detection, lane detection, and path planning. These models help cars
understand their environment, make decisions, and safely navigate
roads.

Sensor Fusion: ML algorithms in MATLAB can combine data from

multiple sensors (LiDAR, radar, cameras) to create more accurate and
reliable models of the surrounding environment.

3. Finance and Stock Market Prediction

42
Algorithmic Trading: ML algorithms in MATLAB can be trained on
historical market data to predict stock prices, forecast trends, and
design trading strategies that maximize returns.

Risk Management: ML techniques such as classification and

regression can be applied to assess financial risks, detect fraud, and
optimize portfolios based on risk-reward trade-offs.

4. Speech Recognition and Natural Language Processing (NLP)

Voice Assistants: MATLAB provides tools for speech recognition and

NLP, enabling the development of voice-controlled systems. Machine
learning models can recognize and transcribe speech or convert text to
speech for applications in customer service, healthcare, and more.

Sentiment Analysis: NLP algorithms can analyze social media or

customer reviews to determine sentiment, helping companies gauge
public opinion and improve customer service.

MATLAB FOR MODEL DEPLOYMENT AND AUTOMATION

43
Matlab is a powerful platform commonly used for model
development, testing, and deployment in various industries, including
engineering, finance, and healthcare. When it comes to model
deployment and applications, Matlab provides several tools and
features that help take models from research and simulation to real-
world use. Here’s a breakdown of how Matlab supports model
deployment and its applications:

1. Matlab Compiler and Compiler SDK

Matlab Compiler allows you to package Matlab code into standalone

applications or shared libraries (.dll, .so, .dylib), making it easy to
deploy models to machines without requiring a full Matlab
installation.

Matlab Compiler SDK enables you to create and deploy Matlab-based

code as callable functions in other programming languages like Java,
.NET, and Python. This is useful for integrating Matlab models into
existing enterprise applications or custom software solutions.

2. MATLAB Production Server

This tool is designed for deploying Matlab functions, applications,

and models to production environments. It allows you to run Matlab
code as a service, enabling access to model predictions or
computations via APIs. This is ideal for integrating predictive models
into web, mobile, or cloud applications.

44
3. Simulink and Simulink Coder

Simulink is a graphical environment for modeling, simulating, and

analyzing dynamic systems. Once a model is developed in Simulink,
Simulink Coder can generate code in C, C++, or other languages,
allowing for deployment on embedded systems and hardware (e.g.,
microcontrollers, FPGA).

This is especially useful for applications in areas like automotive,

aerospace, robotics, and industrial automation where real-time
deployment is required.

4. MATLAB on Cloud Platforms

Matlab integrates with cloud platforms like Amazon Web Services

(AWS), Microsoft Azure, and Google Cloud. You can use Matlab
Parallel Server to distribute and parallelize workloads in the cloud,
and deploy models to handle large-scale or on-demand computation.
Matlab also supports cloud-based data storage and machine learning
workflows.

CHALLENGES AND FUTURE DIRECTIONS

45
Challenges:

1. Scalability with Big Data:

MATLAB can struggle with processing large datasets, especially in
deep learning, due to memory and computational constraints. Though
tools like the Parallel Computing Toolbox help, it still lags behind
frameworks like Apache Spark.

2. Real-Time Processing and Deployment:

MATLAB may not be optimal for real-time, low-latency applications
or resource-constrained devices (e.g., IoT). Although it supports code
generation for hardware, deployment in such systems is often less
efficient compared to alternatives like Python-based tools.

3. Limited Ecosystem for Cutting-Edge ML Techniques:

MATLAB supports many ML models, but it may lag behind other
libraries (e.g., TensorFlow, PyTorch) in supporting the latest
developments, especially in deep learning and reinforcement learning.

4. Cost and Licensing:

The cost of MATLAB licenses can be a barrier for individuals or
small organizations, especially when compared to free, open-source
alternatives.

5. Community Support:

46
While MATLAB has a dedicated user base, its community-driven
resources (forums, tutorials, etc.) are not as vast as those in the Python
ecosystem, limiting learning and support opportunities.

Future Directions:

1. Integration with Open-Source Frameworks:

MATLAB is enhancing support for TensorFlow and PyTorch,
allowing users to combine MATLAB's powerful analysis tools with
the flexibility of open-source ML libraries.

2. Improved Edge and IoT Support:

MATLAB is expanding support for edge devices (e.g., Raspberry Pi,
Jetson). Future developments will likely focus on real-time, energy-
efficient deployments for IoT and embedded systems.
3. Automated Machine Learning (AutoML):
MATLAB is likely to enhance its AutoML tools, enabling users to
automatically select, tune, and deploy machine learning models with
minimal expertise.
4. Cloud-Based ML Solutions:

47
MATLAB will likely expand its cloud offerings, enabling large-scale
training and distributed computing. Integration with platforms like
AWS and Azure could further streamline deployment and
collaboration.

5. Explainability and Fairness:

As AI models are increasingly deployed in critical sectors, MATLAB

will likely focus on enhancing tools for model interpretability,
fairness, and bias detection to ensure ethical AI usage.

48
 CONCLUSION

This report has explored the integration of Machine Learning techniques with
MATLAB, highlighting its powerful capabilities for data analysis, model
building, and deployment. MATLAB provides a comprehensive suite of tools
and functions that make it an ideal environment for implementing machine
learning algorithms, offering flexibility for both beginners and advanced
practitioners. The ease of use of MATLAB's interface, coupled with its
extensive documentation and built-in toolboxes, enables users to efficiently
develop and fine-tune models. Additionally, the report has demonstrated the key
applications of machine learning in various fields, such as pattern recognition,
regression, and classification, with real-world examples to illustrate the
practical value of these methods. Ultimately, MATLAB serves as a robust
platform for accelerating machine learning workflows, providing both
accessibility and advanced functionality, which makes it a valuable tool for
researchers, engineers, and data scientists alike.

REFERENCES

49
 1.www.mathworks.com
2.www.researchgate.net
3.MATLAB for Beginners,Md. Irfan Ali,3rd edition
4.MATLAB Programming for Engineers,Stephen J. Chapman,6th edition
5.MACHINE LEARNING,S.Sridar,4th edition

50