Unit 1

Machine Learning (ML) Tutorial

Machine Learning, often abbreviated as ML is a branch of Artificial Intelligence (AI) that works on algorithm developments and statistical models that allow computers to learn from data and make predictions or decisions without being explicitly programmed. Hence, in simpler terms, machine learning allows computers to learn from data and make decisions or predictions without being explicitly programmed to do so. Essentially, machine learning algorithms learn patterns and relationships from data, allowing them to generalize from instances and make predictions or conclusions on new and uncovered data.

How does Machine Learning Work?

Broadly Machine Learning process includes Project Setup, Data Preparation, Modeling and Deployment. The following figure demonstrates the common working process of Machine Learning. It follows some set of steps to do the task; a sequential process of its workflow is as follows -

Stages of Machine Learning

A detailed sequential process of Machine Learning includes some set of steps of phases which are as –

 

 

 

1.            Data collection: Data collection is an initial step in the process of machine learning. Data is a fundamental part of machine learning, the quality and quantity of your data can have direct consequences for model performance. Different sources such as databases, text files, pictures, sound files, or web scraping may be used for data collection. Data needs to be prepared for machine learning once it has been collected. This process is to organize the data in an appropriate format, such as a CSV file or database, and make sure that they are useful for solving your problem.

2.            Data pre-processing: Pre-processing of data is a key step in the process of machine learning. It involves deleting duplicate data, fixing errors, managing missing data either by eliminating or filling it in, and adjusting and formatting the data. Pre-processing improves the quality of your data and ensures that your machine-learning model can read it right. The accuracy of your model may be significantly improved by this step.

3.            Choosing the right model: The next step is to select a machine learning model; once data is prepared then we apply it to ML Models like Linear regression, decision trees, and Neural Networks that may be selected to implement. The selection of the model generally depends on what kind of data you're dealing with and your problem. The size and type of data, complexity, and computational resources should be taken into account when choosing a model to apply.

4.            Training the model: The next step is to train it with the data that has been prepared after you have chosen a model. Training is about connecting the data to the model and enabling it to adjust its parameters to predict output more accurately. Overfitting and underfitting must be avoided during the training.

5.            Evaluating the model: It is important to assess the model's performance before deployment as soon as a model has been trained. This means that the model has to be tested on new data that they haven't been able to see during training. Accuracy in classifying problems, precision and recall for binary classification problems, as well as mean error squared with regression problems, are common metrics to evaluate the performance of a model.

6.            Hyperparameter tuning and optimization: You may need to adjust its hyperparameters to make it more efficient after you've evaluated the model. Grid searches, where you try different combinations of parameters, and cross-validation, where you divide your data into subsets and train your model on each subset, to ensure that it performs well on different data sets, are techniques for hyperparameter tuning.

7.            Predictions and deployment: As soon as the model has been programmed and optimized, it will be ready to estimate new data. This is done by adding new data to the model and using its output for decision-making or other analysis. The deployment of this model involves its integration into a production environment where it is capable of processing real-world data and providing timely information

Types of Machine Learning

Machine learning models fall into the following categories:

1.            Supervised Machine Learning (SVM): Supervised machine learning uses labeled datasets to train algorithms to classify data or predict outcomes. As input data is inputted into the model, its weights modify until it fits into the model; this process is known as cross validation which ensures the model is not overfitted or underfitted.

Supervised learning helps organizations scale real-world challenges like spam classification in a different folder from your inbox. Different methods for supervised learning include neural networks, naïve Bayes, linear regression, logistic regression, random forest, and SVM.

2.            Unsupervised Machine Learning: Unsupervised machine learning analyses and clusters unlabelled datasets using machine learning methods. The algorithms find hidden patterns or data groupings without human interaction. This method is useful for exploratory data analysis, cross-selling, consumer segmentation, and image and pattern recognition.

It also reduces model features through dimensionality reduction using prominent methods of Principal component analysis (PCA) and singular value decomposition (SVD). Neural networks, k-means clustering, and probabilistic clustering are some popular methods of unsupervised learning.

3.            Semi-supervised learning: As its name implies; Semi-supervised learning is an integration of supervised and unsupervised learning. This method uses both labeled and unlabelled data to train ML models for classification and regression tasks. Semi-supervised learning is a best practice to utilize to solve the problem where a user doesn't have enough labeled data for a supervised learning algorithm.

Hence, it's an appropriate method to solve the problem where data is partially labeled or unlabelled. Self-training, co-training, and graph-based labeling are some of the popular Semi-supervised learning methods.

4.            Reinforcement Machine Learning: Reinforcement machine learning is a type of machine learning model that is similar to supervised learning but does not use sample data to train the algorithm. This model learns by trial and error.

A series of good results will be reinforced to create the optimal proposal or policy for a specific problem.

Common Machine Learning Algorithms

Several machine learning algorithms are commonly used. These include:

·               Neural networks: Neural networks function similarly to the human brain, comprising multiple linked processing nodes. Neural networks excel at pattern identification and are used in different applications such as natural language processing, image recognition, speech recognition, and creating images.

·               Linear regression: This algorithm predicts numerical values using a linear relationship between variables. For example, linear regression is used to forecast housing prices based on past data in a particular area.

·               Logistic regression: This supervised learning method predicts categorical variables, such as "yes/no" replies to questions. It is suitable for applications such as spam classification and quality control on a production line.

·               Clustering: Clustering algorithms use unsupervised learning to find patterns in data and organise it accordingly. Computers can assist data scientists by identifying differences between data items that humans have overlooked.

·               Decision trees: Decision trees are useful for categorising data and for regression analysis, which predicts numerical values. A tree structure can be used to illustrate the branching sequence of linked decisions used in decision trees. Unlike neural networks, decision trees can be easily validated and audited.

·               Random forests: ML predicts a value or category by integrating results from different decision trees.

Importance of Machine Learning

Machine Learning is important in automation, extracting insights from data, and decision-making processes. It has its significance due to the following reasons:

·               Data processing: The main reason machine learning has become so important is to process and make sense of large amounts of data. Traditional methods of data analysis are becoming insufficient given the explosion of digital information coming from social media, sensors, and other sources. This data is important and reveals hidden patterns and provides invaluable insight for decision-making processes, which can be exploited by machine learning algorithms.

·               Data-driven insights: Machine learning algorithms can find patterns, trends, and correlations in big data sets that humans cannot. Better decisions and forecasts can be made with this information.

·               Automation: Machine learning automates manual activities, saving time and decreasing errors by learning from data and improving over time, ML algorithms can perform previously manual tasks, freeing humans to focus on more complex and creative tasks. This not only increases efficiency but also opens up new possibilities for innovation. Data entry, classification, and anomaly detection can be automated with machine learning.

·               Personalization: User preferences and behavior can be analyzed using machine learning algorithms to generate personalized recommendations and experiences. It is most widely used in social media like e-commerce, and streaming services by providing a way to increase user engagement and satisfaction.

·               Predictive analytics: Models of machine learning may be trained to predict subsequent outcomes based on past data. This is useful for different applications like sales forecasts, risk management, and demand planning.

·               Optimization: Machine learning algorithms optimize systems and processes for efficiency and performance. Their smart grid optimizations include supply chain logistics, resource allocation, and energy consumption.

·               Pattern recognition: Machine learning is useful in image, audio, and natural language processing because it can recognize complicated data patterns easily and timely.

·               Healthcare: Machine learning is used in disease diagnosis, outbreaks; personalized patient treatment plans, personalized treatment planning, medical imaging accuracy, and drug discovery. It accurate diagnosis, medical image processing, genomic data, and electronic health records.

·               Finance: Machine learning is used for credit scoring, algorithmic trading, and fraud detection.

·               Retail: Machine learning can also be used for recommendation systems, supply chains, or customer service.

·               Fraud detection and cybersecurity: Machine Learning algorithms can detect patterns of fraudulent behavior for financial transactions by detecting and mitigating security threats in real-time, it is used to enhance cybersecurity as well.

·               Continuous improvement: It is possible to train and update machine learning models with new data at regular intervals, enabling them to adapt to changes in the environment as well as improve over time.

Machine Learning enables organizations to take advantage of the power of data to gain insight, streamline processes and drive innovation throughout a variety of sectors.

Applications of Machine Learning

Nowadays; Machine Learning is used almost everywhere. However, some most commonly used applicable areas of Machine Learning are:

·               Speech recognition: It is also known as automatic speech recognition (ASR), computer speech recognition, or speech-to-text, and it is a capability that uses natural language processing (NLP) to translate human speech into a written format. To perform voice search, such as Siri, or improve text accessibility, a large number of Mobile Devices incorporate speech recognition into their systems.

·               Customer service: Chatbots are replacing human operators on websites and social media, affecting client engagement. Chatbots answer shipping FAQs, offer personalized advice, cross-sell products, and recommend sizes. Some common examples are virtual agents on e-commerce sites, Slack and Facebook Messenger bots, and virtual and voice assistants.

·               Computer vision: This artificial intelligence technology allows computers to derive meaningful information from digital images, videos, and other visual inputs that can then be used for appropriate action. Computer vision, powered by convolutional neural networks, is used for photo tagging on social media, radiology imaging in healthcare, and self-driving cars in the automotive industry.

·               Recommendation engines: AI algorithms may help to detect trends in data that might be useful for developing more efficient marketing strategies using past data patterns. Online retailers use recommendation engines to provide their customers with relevant product recommendations for the purchasing process.

·               Robotic process automation (RPA): Also known as software robotics, RPA uses intelligent automation technologies to perform repetitive manual tasks.

·               Automated stock trading: AI-driven high-frequency trading platforms are designed to optimize stock portfolios and make thousands or even millions of trades each day without human intervention.

·               Fraud detection: Machine learning is capable of detecting suspected transactions for banks and others in the financial sector. A model can be trained by supervised learning, based on knowledge of recent fraudulent transactions. Anomaly detection may identify transactions that appear unusual, and need to be followed up.

Target Audience

This machine learning tutorial has been prepared for those who want to learn about the basics and advances of Machine Learning. In a broader sense; ML is a subset of Artificial Intelligence (AI) that focuses on developing algorithms and models that allow computers to learn from data and make predictions or decisions without being explicitly programmed to do so. Machine learning requires data. This data can be text, images, audio, numbers, or video. The quality and quantity of data considerably affect machine learning model performance. Features are data qualities used to predict or decide. Feature selection and engineering entail selecting and formatting the most relevant features for the model.

Prerequisites to Learn Machine Learning

You should have a basic understanding of the technical aspects of Machine Learning. Learners should be familiar with data, information, and its basics. Knowledge of Data, information, structured data, unstructured data, semi-structured data, data processing, and Artificial Intelligence basics; Proficiency in labeled / unlabelled data, feature extraction from data, and their application in ML to solve common problems is a must.

Algorithms and mathematical models are the most essential things to learn before exploring Machine Learning concepts. These prerequisites give a solid basis for Machine Learning, but it's also important to understand that the specific requirements may vary as per Machine Learning models, complexity, cutting-edge technologies, and nature of the work.

.

Need for Machine Learning

Human beings, at this moment, are the most intelligent and advanced species on earth because they can think, evaluate and solve complex problems. On the other side, AI is still in its initial stage and hasn’t surpassed human intelligence in many aspects.

Then the question is, what is the need to make machines learn? The most suitable reason for doing this is “to make decisions, based on data, with efficiency and scale”.

Lately, organizations are investing heavily in newer technologies like Artificial Intelligence, Machine Learning and Deep Learning to get the key information from data to perform several real-world tasks and solve problems. We can call it data-driven decisions taken by machines, particularly to automate the process.

These data-driven decisions can be used, instead of programming logic, in problems that cannot be programmed inherently. The fact is that we can’t do without human intelligence, but another aspect is that we all need to solve real-world problems with efficiency at a huge scale. That is why the need for machine learning arises.

History of Machine Learning

The history of Machine learning roots back to the year 1959, when Arthur Samuel invented a program that calculates the winning probability in checkers for each side.

Well, the evolution of Machine learning through decades started with the question, "Can Machines think?". Then came the rise of neural networks between 1960 and 1970. Machine learning continued to advance through statistical methods such as Bayesian networks and decision tree learning.

The revolution of Deep Learning started off in the 2010s with the evolution of tasks such  as natural language processing, convolution neural networks and speech recognition. Today, machine learning has turned out to be a revolutionizing technology that has become a part of all fields, ranging from healthcare to finance and transportation.

Machine Learning Algorithms Vs. Traditional Programming

The difference between machine algorithms and traditional programming depends on how they are programmed to handle tasks. Some comparisons based on different criteria are tabulated below:

Criteria	Machine learning algorithms	Traditional programming
Problem solving approach	The computer learns from training a model on large datasets.	Explicit rules are given to the computer to follow in the form of code that is manually programmed.
Data	They heavily rely on data, it defines the performance of the model.	They rely less on data, as the output depends on the logic encoded.
Complexity of Problem	Best suited for complex problems like image segmentation or natural language processing, which require identifying patterns and relationships in the data.	Best suited for a problem with defined outcome and logic.
Flexibility	It is highly flexible and adapts to different scenarios, especially because the model is retrained with new data.	It has limited flexibility, as the changes should be done manually.
Outcome	The outcome in machine learning is unpredictable, as it depends on data trained, model and many other things.	The outcome in traditional programming can be accurately predicted if the problem and logic are known.

Machine Learning Vs. Deep Learning

Deep learning is a sub-field of Machine learning. The actual difference between these is the way the algorithm learns.

In Machine learning, computers learn from large datasets using algorithms to perform tasks like prediction and recommendation. Whereas Deep learning uses a complex structure of algorithms developed similar to the human brain.

The effectiveness of deep learning models for complex problems is more compared to machine learning models. For example, autonomous vehicles are usually developed using deep learning where it can identify a U-TURN sign board using image segmentation while if a machine learning model was used, the features of the signboard are selected and then identified using a classifier algorithm.

Machine Learning Vs. Generative AI

Machine learning and Generative AI are different branches with different applications. While Machine Learning is used for predictive analysis and decision-making, Generative AI focuses on creating content, including realistic images and videos in existing patterns.

Future of Machine Learning

Machine Learning is definitely going to be the next game changer in technology. Automated machine learning and synthetic data generation, are new age developments that make machine learning more accessible and efficient.

One big technology that is an adoption of machine learning is Quantum computing. It uses the mechanical phenomenon of quantum to create a system that exhibits multiple states at the same time. These advanced quantum algorithms are used to process data at high speed. AutoML is another technology that combines automation and machine learning. It potentially includes each stage from raw data to developing a model ready for deployment.

Multi-modal AI is an AI system used to effectively interpret and analyze multi-sensory inputs, including texts, speech, images, and sensor data. Generative AI is another emerging application of machine learning which focuses on creating new content that mimics existing patterns. A few other emerging technologies that have an impact on Machine learning are Edge computing, Robotics, and many more.

Training

Training is the process of teaching the machine learning algorithm to make accurate predictions or decisions. This is done by providing the algorithm with a large dataset and allowing it to learn from the patterns and relationships in the data. During training, the algorithm adjusts its internal parameters to minimize the difference between its predicted output and the actual output.

Testing

Testing is the process of evaluating the performance of the machine learning algorithm on a separate dataset that it has not seen before. The goal is to determine how well the algorithm generalizes to new, unseen data. If the algorithm performs well on the testing dataset, it is considered to be a successful model.

Overfitting

Overfitting occurs when a machine learning model is too complex and fits the training data too closely. This can lead to poor performance on new, unseen data because the model is too specialized to the training dataset. To prevent overfitting, it is important to use a validation dataset to evaluate the model's performance and to use regularization techniques to simplify the model.

Underfitting

Underfitting occurs when a machine learning model is too simple and cannot capture the patterns and relationships in the data. This can lead to poor performance on both the training and testing datasets. To prevent underfitting, we can use several techniques such as increasing model complexity, collect more data, reduce regularization, and feature engineering.

It is important to note that preventing underfitting is a balancing act between model complexity and the amount of data available. Increasing model complexity can help prevent underfitting, but if there is not enough data to support the increased complexity, overfitting may occur instead. Therefore, it is important to monitor the model's performance and adjust the complexity as necessary.

Machine Learning Model

Before discussing the machine learning model, we must need to understand the following formal definition of ML given by professor Mitchell −

“A computer program is said to learn from experience E with respect to some class of tasks T and performance measure P, if its performance at tasks in T, as measured by P, improves with experience E.”

The above definition is basically focusing on three parameters, also the main components of any learning algorithm, namely Task(T), Performance(P) and experience (E). In this context, we can simplify this definition as −

ML is a field of AI consisting of learning algorithms that −

·         Improve their performance (P)

·         At executing some task (T)

·         Over time with experience (E)

Based on the above, the following diagram represents a Machine Learning Model −

Let us discuss them more in detail now −

Task(T)

From the perspective of problem, we may define the task T as the real-world problem to be solved. The problem can be anything like finding best house price in a specific location or to find best marketing strategy etc. On the other hand, if we talk about machine learning, the definition of task is different because it is difficult to solve ML based tasks by conventional programming approach.

A task T is said to be a ML based task when it is based on the process and the system must follow for operating on data points. The examples of ML based tasks are Classification, Regression, Structured annotation, Clustering, Transcription etc.

Experience (E)

As name suggests, it is the knowledge gained from data points provided to the algorithm or model. Once provided with the dataset, the model will run iteratively and will learn some inherent pattern. The learning thus acquired is called experience(E). Making an analogy with human learning, we can think of this situation as in which a human being is learning or gaining some experience from various attributes like situation, relationships etc. Supervised, unsupervised and reinforcement learning are some ways to learn or gain experience. The experience gained by out ML model or algorithm will be used to solve the task T.

Performance (P)

An ML algorithm is supposed to perform task and gain experience with the passage of time. The measure which tells whether ML algorithm is performing as per expectation or not is its performance (P). P is basically a quantitative metric that tells how a model is performing the task, T, using its experience, E. There are many metrics that help to understand the ML performance, such as accuracy score, F1 score, confusion matrix, precision, recall, sensitivity etc.