[Analysis] Types of Machine Learning – Supervised | Unsupervised | Semi-supervised | Reinforcement

  • Blog|FEMA & Banking|
  • 8 Min Read

Supervised, unsupervised, semi-supervised and reinforcement learning explained from IIBF’s Emerging Technologies courseware, with the differences between them and how banks use each type.

  • By Taxmann
  • |
  • Last Updated on 14 November, 2026

Types of Machine Learning

Machine learning has four types: supervised, unsupervised, semi-supervised and reinforcement learning. The IIBF courseware on Emerging Technologies sorts the types of machine learning by learning style. Supervised learning trains on labelled data to predict a category or a number. Unsupervised learning finds hidden patterns in unlabelled data. Semi-supervised learning combines a small labelled dataset with a larger unlabelled one. Reinforcement learning learns by trial and error, through rewards and penalties. In banking, the book links them to fraud detection, credit risk scores and trading decisions.

Table of Contents

  1. What is machine learning?
  2. What are the four types of machine learning?
  3. What is supervised learning?
  4. What is unsupervised learning?
  5. What is the difference between supervised and unsupervised learning?
  6. What is semi-supervised learning?
  7. How does reinforcement learning work?
  8. How do banks use machine learning?
  9. Quick answers

1. What is machine learning?

Machine learning lets computers learn from data without being explicitly programmed. It is a sub-field of artificial intelligence in which an algorithm is fed a large volume of data, finds patterns in it and uses them to predict future outcomes and make data-driven recommendations. Our post on the types of artificial intelligence shows where it sits within AI.

The difference from ordinary programming lies in who writes the rules. In a regular program, the programmer tells the computer what to do, and the flow and the outcome are clearly defined. In machine learning, the machine is given the source data and the expected outcome, and its task is to find the patterns, or rules, that link the two. The algorithms are relatively simple; in the book’s words, “the complexity of the structure lies in the data”.

This is why machine learning goes hand in hand with big data. Humans cannot work through such large volumes of data quickly enough to find patterns and make predictions. A machine learning model can, and it keeps adapting to new data and experience to improve its performance over time.

2. What are the four types of machine learning?

Machine learning is grouped into four types, based on the learning style and method adopted: supervised, unsupervised, semi-supervised and reinforcement learning. The quickest way to tell them apart is to look at the data each one learns from, and at how it finds out whether its answer was right.

Type Training data How it learns Typical tasks Banking example
Supervised learning Labelled data: each input comes with the desired output Learns the mapping from input to output, then predicts outputs for new data Classification (a category) and regression (a number) Fraud detection and risk assessment; finding card customers at risk of leaving
Unsupervised learning Unlabelled data Discovers hidden patterns, groups and associations on its own Clustering, association and anomaly detection Anomaly detection to find fraudulent transactions
Semi-supervised learning A small labelled dataset and a larger unlabelled one Clusters the large dataset, then applies the labels learned from the small one Problems where good-quality labelled data is scarce None given
Reinforcement learning No preset training dataset; feedback from the environment Trial and error, with decisions rewarded or punished Sequential decisions such as pricing, delivery routes and trading Choosing trading strategies; deciding whether to hold, buy or sell

Taxmann.com | Research | Subscribe Now!

3. What is supervised learning?

Supervised learning trains a model on labelled data, so that it learns to give the right output for a new input. As the name suggests, it needs external supervision. Each training example pairs an input with the desired output. The algorithm uses these known outputs to build an equation for the label or value, a step known as the training stage. It then compares its results with the correct ones, measures the errors and improves. The aim is to generalise from known data to data the model has not seen.

Supervised learning, also called inductive learning, is the most developed and mature form of learning and the most widely used. It is among the most popular methods for prediction, fraud detection and risk assessment. It does, however, require precise labelling of data.

Supervised algorithms fall into two groups, according to the output they produce:

  • Classification assigns data to predefined classes, such as yes or no, 0 or 1, or spam or not spam. For example, an algorithm trained on images of tomatoes and apples can tell which of the two a new image shows.
  • Regression predicts a continuous numerical value, such as the price of a property, stock price movements or business growth. A model trained on houses whose prices, locations, sizes and amenities are known can estimate the price of a newly built house.

A supervised learning project moves through data collection, data preparation, modelling, model evaluation, deployment and monitoring. Its use cases include spam detection, customer sentiment analysis, image and object recognition, and predictive analytics systems that help a business forecast outcomes.

4. What is unsupervised learning?

Unsupervised learning trains a model on unlabelled data and lets the algorithm find the patterns itself. Nobody tells it what to look for. It discovers hidden patterns, similarities and relationships without any added bias from the user, which makes it useful for exploratory analysis of data.

Its common uses are:

  • Clustering, which splits a dataset into groups so that items in one group are more like each other than like items in other groups. Customer segmentation for marketing and business strategy is an example.
  • Association, which identifies items that often occur together and predicts rules that describe the data.
  • Anomaly detection, which picks out unusual data points. In banking, this is useful for finding fraudulent transactions.
  • Recommender systems, which group users with similar viewing patterns and suggest similar content.

Unlabelled data is easy to obtain. A large collection of customers’ comments from social media and e-mail, for instance, can be sorted by an unsupervised algorithm into categories nobody defined, which helps a business see the topics being discussed and the sentiment about a product. The trade-off is control. The outcome cannot be predicted in advance, and the method works much like a black box: most of the input and output steps are not controllable by the user.

5. What is the difference between supervised and unsupervised learning?

The difference is the label. Supervised learning learns from data that already carries the right answers, while unsupervised learning works on data with no answers attached. The other differences follow from that one.

  • Supervised learning needs external supervision and precise labelling. Making a dataset machine-readable for unsupervised learning needs no additional human effort, so it can work on much larger datasets.
  • A supervised model predicts a known kind of output, a category or a number. An unsupervised model discovers groups and rules that nobody set out in advance.
  • Supervised learning is the most mature and widely used form. Unsupervised learning services are less popular, mainly because they have fewer applications in everyday life, though unsupervised learning is often used in the more advanced applications of artificial intelligence.

Fraud detection draws on both. Banks build fraud prediction models on past data, while anomaly detection, an unsupervised method, flags transactions that look unlike the rest, working on unlabelled data.

6. What is semi-supervised learning?

Semi-supervised learning sits between the two. It trains on a small labelled dataset together with a larger unlabelled one. The algorithm first clusters the large dataset and then applies the labels learned from the small one. It is often used when high-quality labelled data is lacking, and it reduces the dependence on large labelled datasets. Because neither supervised nor unsupervised algorithms can make good use of such a mix of labelled and unlabelled data, it needs specialised algorithms of its own.

7. How does reinforcement learning work?

Reinforcement learning learns by doing. A component called an agent acts in an environment, receives feedback on each action and improves through trial and error, with a reward and penalty system guiding it towards the best option. Supervised learning trains with an answer key. Reinforcement learning has none, so the agent has to learn from its own mistakes.

Five features set reinforcement learning apart:

  • there is no supervisor, only a reward;
  • decisions are made sequentially;
  • time matters;
  • feedback is delayed, not instant; and
  • the agent decides the next data it receives.

Reinforcement learning works best when the application scenario and the best and worst possible outcomes are known. Its applications include dynamic pricing, where Q-learning is used to build pricing models; robotics in industrial automation; simulators for training pilots and train drivers; game logic; last-mile delivery routes; and financial investment decisions on the best trading strategies. In trading, an agent can decide whether to hold, buy or sell, and the model is evaluated against market benchmarks to check that it works properly.

8. How do banks use machine learning?

Banks use machine learning where they hold large volumes of data, and financial services produce exactly that: data on daily transactions, invoices, payments, vendors and consumers. The IIBF courseware sums up the gains as greater revenue from better productivity and user experience, lower operating costs through automation, and stronger security leading to better compliance.

Several banking use cases point to a specific type of learning:

Banking task How it works Type of learning
Keeping high-value credit card customers Binary classification models identify consumers at risk; a recommender model then picks the best card offers to keep them Supervised (classification), followed by a recommender model
Catching fraudulent transactions Anomaly detection finds unusual data points in a dataset Unsupervised
Filtering phishing e-mails Phishing e-mails are identified and filtered with classification and regression algorithms Supervised
Choosing trading strategies An agent decides whether to hold, buy or sell, and is evaluated against market benchmarks Reinforcement

Machine learning also helps in the prevention of money laundering, in risk scores for consumers seeking loans, and in real-time approvals of transactions with fewer incorrect denials. On the customer side, Indian banks have deployed chatbots such as SBI Intelligent Assistant (SIA), which can respond in 14 languages by speech or text, EVA of HDFC Bank and IPal by ICICI Bank. Our post on AI in banking covers chatbots and credit decisions in more detail.

9. Quick answers

What are the four types of machine learning?

Supervised, unsupervised, semi-supervised and reinforcement learning. They are grouped by learning style: whether the model learns from labelled data, unlabelled data, a mix of the two, or rewards and penalties.

What is the difference between supervised and unsupervised learning?

Supervised learning trains on labelled data and predicts a known kind of output, such as a category or a number. Unsupervised learning trains on unlabelled data and discovers hidden patterns, groups and associations on its own.

What is supervised learning in machine learning?

It is learning from labelled examples, where each input comes with the desired output. The model learns the mapping and uses it to predict outputs for new data, through classification or regression.

What is unsupervised learning?

It is learning from unlabelled data. The algorithm finds hidden patterns by itself and is used for clustering, association and anomaly detection, such as finding fraudulent transactions.

What is semi-supervised learning?

It trains on a small labelled dataset and a larger unlabelled one. The algorithm clusters the large dataset and then applies the labels learned from the small one, which helps when high-quality labelled data is lacking.

What is reinforcement learning?

It is learning by trial and error. An agent acts in an environment and its decisions are rewarded or punished, so it learns the best course of action without a preset training dataset.

Which type of machine learning is used for fraud detection in banks?

Both main types. Supervised learning is among the most popular methods for fraud detection, and unsupervised anomaly detection is useful for finding fraudulent transactions in banking.

Is machine learning a part of artificial intelligence?

Yes. Machine learning is a sub-field of artificial intelligence that learns from data instead of following explicitly programmed instructions.


Chapter 4 of Emerging Technologies (2026 Edition), the Indian Institute of Banking & Finance’s courseware for its Certificate Course on Emerging Technologies, published by Taxmann, covers the types of machine learning, the stages of machine learning, regression, classification and clustering algorithms, association rules, and machine learning adoption and use cases in banks. It ends with multiple choice questions and terminal questions for practice.

For the rules that banks work under, Taxmann.com | Research, FEMA & Banking brings together the Acts, Rules, circulars and notifications, case laws and articles on FEMA, banking and insurance. Taxmann.com | Store also lists IIBF’s courseware for its other certificate courses, including Information System for Banks and IT Security.

Related reading: RBI’s draft guidance on model risk management · SEBI’s AI/ML framework for the securities market.

More in this series: emerging technologies in banking · types of artificial intelligence · overfitting and underfitting · types of data analytics · AI in banking.

Based on the 2026 Edition of Emerging Technologies, IIBF’s courseware for the Certificate Course on Emerging Technologies.

Disclaimer: The content/information published on the website is only for general information of the user and shall not be construed as legal advice. While the Taxmann has exercised reasonable efforts to ensure the veracity of information/content published, Taxmann shall be under no liability in any manner whatsoever for incorrect information, if any.

Leave a Reply

Your email address will not be published. Required fields are marked *

Everything on Tax and Corporate Laws of India

To subscribe to our weekly newsletter please log in/register on Taxmann.com

Author: Taxmann

Taxmann Publications has a dedicated in-house Research & Editorial Team. This team consists of a team of Chartered Accountants, Company Secretaries, and Lawyers. This team works under the guidance and supervision of editor-in-chief Mr Rakesh Bhargava.

The Research and Editorial Team is responsible for developing reliable and accurate content for the readers. The team follows the six-sigma approach to achieve the benchmark of zero error in its publications and research platforms. The team ensures that the following publication guidelines are thoroughly followed while developing the content:

  • The statutory material is obtained only from the authorized and reliable sources
  • All the latest developments in the judicial and legislative fields are covered
  • Prepare the analytical write-ups on current, controversial, and important issues to help the readers to understand the concept and its implications
  • Every content published by Taxmann is complete, accurate and lucid
  • All evidence-based statements are supported with proper reference to Section, Circular No., Notification No. or citations
  • The golden rules of grammar, style and consistency are thoroughly followed
  • Font and size that's easy to read and remain consistent across all imprint and digital publications are applied