Machine Learning for Recommender Systems: A Beginner's Guide

Learn how to use Amazon, Netflix, Facebook and LinkedIn's recommender technologies to influence and increase sales
3.9 (48 ratings)
Instead of using a simple lifetime average, Udemy calculates a
course's star rating by considering a number of different factors
such as the number of ratings, the age of ratings, and the
likelihood of fraudulent ratings.
1,211 students enrolled
92% off
Take This Course
  • Lectures 23
  • Length 1 hour
  • Skill Level All Levels
  • Languages English, captions
  • Includes Lifetime access
    30 day money back guarantee!
    Available on iOS and Android
    Certificate of Completion
Wishlisted Wishlist

How taking a course works


Find online courses made by experts from around the world.


Take your courses with you and learn anywhere, anytime.


Learn and practice real-world skills and achieve your goals.

About This Course

Published 3/2016 English Closed captions available

Course Description

Have you ever wondered:

  1. How does Amazon recommend products you might be interested in purchasing? OR
  2. How does Netflix decide which movies or TV shows you might want to watch? OR
  3. How does Facebook or LinkedIn decide who might you want to form a link with? OR
  4. How does Udemy decide what courses to market to you? OR
  5. How does New York Times decide which news you might be interested in reading?

If you have and you want to learn the science behind them, you have come to the right place. In this course, I will show you how these companies use Recommender systems or Machine Learning to influence your purchasing decisions. This course is timely and extremely relevant now as almost all major service-oriented companies function on recommender systems.

You will understand how these systems work and learn how to build and use your own recommender systems, just like these big companies do.

Learn how to build the recommender systems that are being used by almost every big service-oriented company in today’s world with this introductory course for beginners.

  • Goals and applications of recommender systems
  • News recommendation, products you may like and movie suggestions
  • Popularity-based systems, Collaborative Filtering and Co-occurrence matrix
  • Matrix Factorization and estimated Topic Vectors
  • Cold-start problem and how to handle it
  • Precision, Recall and Optimal recommenders
  • Free 11000-word e-book on recommender systems

Recommender Systems have changed the way people find products, information, and even other people. They study patterns of behavior to know what someone will prefer from among a collection of things they have never experienced. The technology behind recommender systems has evolved over the past 20 years into a rich collection of tools that enable the practitioner or researcher to develop effective recommenders. Such systems are being used by companies such as Amazon, Facebook, Netflix, LinkedIn, Quora, Udemy, New York Times, etc. By taking this course, you will learn the most important of those tools, including how they work, how to use them, how to evaluate them, and their strengths and weaknesses in practice. The algorithms you will study include popularity-based systems, classification-based approach, collaborative filtering, matrix recommendation, etc.

Content and Overview

This course contains 23 lectures, 1 hour of content, 4 quizzes and one 11000 word e-book written by me. It is designed for anyone with an understanding of basic mathematics, who wishes to understand the technology behind the recommender systems they encounter every day. By taking this course you will establish a strong understanding of the concept behind recommender systems.

Starting with properly defining the goals of a recommender system, this course will show you numerous practical examples of where these systems are found in our daily lives. Then you will jump right into the action and start building your first recommender systems in the first section itself.

In the second section, you will move on to design more sophisticated systems and eventually learn how to develop a product recommendation system similar to Amazon. You will learn how to implement Collaborative Filtering, tackle the adverse effects of very popular items, construct and normalize the Co-occurrence matrix and leverage purchase histories for making better recommendations.

In the third section, you will learn how to develop a movie recommendation system like Netflix by Matrix Factorization. You will learn how to automatically construct “topic” vectors corresponding to users and movies that will help you in predicting movie ratings. Moreover, you will discover the infamous cold-start problem and understand how to solve it by blending or combining different recommendation approaches.

After making you familiar with popular recommender systems, the course will show you how to evaluate the performance of these systems by metrics such as precision and recall. You will know about optimal recommenders and learn how to make the best recommender in a constrained practical scenario.

Students completing the course will be literate in one of the most widely used tools of Machine Learning namely Recommender Systems.

They will be able to identify what recommender system is being used by a company whenever it markets an item (product, movie, news, etc.) to them. Moreover, if you want to design a recommender system for your own business or work, you will know by intuition which ones to try and how to evaluate and compare their performance before you select the final one.

The massive 11000-word e-book, which you receive as a part of this course not only contains the ideas taught in the lecture videos, but also has bonus materials that reinforce the concepts you have learned and give you new tricks to design recommender systems.

What are the requirements?

  • At least high school level math skills will be required.

What am I going to get from this course?

  • An 11000-word e-book on recommender systems written by me
  • Describe the goals of a recommender system
  • Provide examples of applications where recommender systems are useful
  • Implement a popularity based recommender system like New York Times.
  • Implement a co-occurrence based product recommendation system like Amazon for your own business
  • Exploit "topic" vectors estimated by matrix factorization for implementing a movie recommendation system like Netflix
  • Describe the cold-start problem and offer ways to handle it (e.g., incorporating features)
  • Analyze performance of various recommender systems in terms of precision and recall

Who is the target audience?

  • Anyone interested in understanding the technology behind the recommendations he/she receives every day from social media, movie streaming sites and e-commerce stores.
  • Machine Learning enthusiasts or data scientists looking to enhance their skill set
  • Entrepreneurs interested in building a recommender system for their own business

What you get with this course?

Not for you? No problem.
30 day money back guarantee.

Forever yours.
Lifetime access.

Learn on the go.
Desktop, iOS and Android.

Get rewarded.
Certificate of completion.


Section 1: Introduction and Overview

In this lecture, I thank and welcome you to this course.

  • First, I help you in deciding whether this course is right for you.
  • Second, I briefly describe what you will learn from this course.
  • Third, I discuss why I am suitable for teaching you this course.
  • Finally, I share my motto with you and show why you are here for a unique journey.

In this lecture, I will show you the structure and organization of this course.

In this lecture, I have provided the slides used in this section.
Section 2: Introduction to Recommender Systems

Recommender systems are one of the most important practical application of Machine Learning and this lecture will welcome you to the world of recommender systems. Moreover, you will see how Amazon and Netflix became the early pioneer of this field.


In this lecture, you will see how widespread the world of Recommender Systems is. You will learn the key element that makes these systems so good - personalization. Specifically, you will find how Recommender Systems are being used companies such as YouTube, Netflix, Amazon, Pandora, Facebook, LinkedIn, etc. Moreover, you will learn that, in recent years, giants from the medical industry are looking into these systems for a specific purpose which will allow them to bypass the FDA approval procedure.


In this lecture, you will learn two popular models of Recommender Systems.

  • First, you will understand the popularity-based model and know how it is used by New York Times. Moreover, you will learn the limitation of this model.
  • Second, you will learn about the classification-based model and understand how it implements personalization. Furthermore, you will know the pros and cons of this model.

In this lecture, I have provided the slides used in this section.

5 questions

In this quiz, you will be asked questions from the basics of Recommender Systems.

Section 3: Collaborative Filtering and Co-occurrence Matrices

In this lecture, you will encounter the third type of recommender system which serves as the basis of Amazon's product recommendations.

  • First, I will discuss the idea of Collaborative Filtering and how to implement that by using Co-occurrence matrix.
  • Second, you will learn how to construct the Co-occurrence matrix and use it for product recommendations.

In this lecture, I will show you a practical problem that Collaborative Filtering faces which you must keep in mind while creating your own co-occurrence based recommender systems.

  • First, I will explain this problem to you and demonstrate it by taking an example.
  • Second, I will show you how adversely this problem affects a co-occurrence based recommender system.

In this lecture, I will show you how you can solve the problem posed by very popular items by normalizing the co-occurrence matrix using Jaccard Similarity. First, I will tell you about the method of normalization and subsequently, I will explain it in a simple manner by using Venn diagrams.


In this lecture, you will learn how to leverage the purchase history of a user to make better recommendations for him or her. Moreover, you will get to know the limitations of Collaborative Filtering and learn about the infamous cold start problem.


In this lecture, I have provided the slides used in this section.

5 questions

In this quiz, you will be asked questions related to Collaborative Filtering and Co-occurrence matrix.

Section 4: Matrix Factorization

In this lecture, you will learn a sophisticated recommendation technique that combines the advantages of the classification model (Method 1) and collaborative filtering (Method 2). This technique is termed as Matrix Factorization and is employed by numerous companies around the globe. However, I will discuss this method in connection with the movie recommendation task since Netflix was the one who made this technique famous.


In this lecture, you will know how to use the known features of movies and users for making recommendations.

  • First, you will understand how to define the movies and the users in terms of vectors.
  • Second, you will learn how to mathematically calculate whether a particular user will like a particular movie.
  • Third, you will understand how to use the above two concepts for recommending movies to users.

In this lecture, you will see how to store the user and movie vectors in their corresponding matrices. Moreover, you will learn the method of generating the Rating matrix by performing matrix multiplication of the above two matrices.


In this lecture, you will learn how to automatically discover users and movies topic vectors from the data and perform predictions

  • First, you will see how the metric Residual Sum of Squares is employed to derive the topic vectors by using the observed ratings.
  • Second, you will understand how the derived parameters can be used to recommend movies.
  • Third, you will know the limitation of Matrix Factorization.

In this lecture, you will learn how to solve the cold start problem faced by Matrix Factorization. The technique you will use is blending different models to create a more powerful model. In this connection, you will see that the winning entry of the 1 million USD Netflix Prize was a model that blended 100 different recommender models.


In this lecture, I have provided the slides used in this section.

5 questions

In this quiz, you will be asked questions on Matrix Factorization.

Section 5: Performance metrics for recommender systems

In this lecture, you learn how to evaluate the performance of recommender systems.

  • First, you will see why the mostly used performance metric for Machine Learning systems i.e. classification accuracy cannot be used in these tasks.
  • Second, I will show you alternative metrics termed as Precision and Recall that are apt of recommender systems.
  • Third, I will take the case of product recommendations and you will learn how to calculate Precision and Recall for an example problem.

In this lecture, I will demonstrate to you that maximizing recall by an improper way might lead to a very low value of precision. Moreover, you will learn the concept of optimal recommenders.


In this lecture, I have provided the slides used in this section.

3 questions

In this quiz, you will be asked questions related to the performance metrics of recommender systems.

Section 6: Conclusion and Bonus materials

In this lecture, I conclude this course and make some parting comments.I have also added the slides in the Resources area, for your reference.


In this lecture, you will found your FREE ebook. Enjoy reading it!

Students Who Viewed This Course Also Viewed

  • Loading
  • Loading
  • Loading

Instructor Biography

Dr. Subhrajit Roy, Scientist by profession. Teacher at heart.

I am a Scientist at IBM Research - Australia, working with the TrueNorth team on Machine Learning and its applications to mobile processors. I completed my Ph.D. from the Nanyang Technological University, Singapore on Neural Networks and Deep Learning. Like most researchers, I juggle between performing research and teaching students. My research involves developing novel algorithms that will be applied to automatically program the IBM's revolutionary neural network computer i.e. the TrueNorth system. I have published over 20 papers in international journals and conferences and you can check out my most cited papers below.

On the other hand, I love to teach my students the things I work on. After getting excellent feedback from the students I teach at the local universities, I decided to cater to the huge online community. My motto is to teach Machine Learning in a simple way such that it does not seem difficult and becomes accessible to everyone. I believe that behind each successful Machine Learning algorithm there is a physical significance or intuition that makes it work. Mathematics is required only to validate it. This thought pushed me into the path of creating and publishing courses related to Machine Learning, Big Data and Neural Networks where I shall discover and share those intuitions with you.

Please don't hesitate to drop me a message if you have a suggestion for a topic for one of my courses, or need help with something. I would love to talk to you.

My selected publications:

1. S. Roy and A. Basu, "An Online Structural Plasticity Rule for Generating Better Reservoirs", Neural Computation, MIT Press, 2016.

2. S. Roy and A. Basu, "An Online Unsupervised Structural Plasticity Algorithm for Spiking Neural Networks," IEEE Transactions on Neural Networks and Learning Systems, 2016.

3. S. Roy, P. P. San, S. Hussain, L. W. Wei and A. Basu, "Learning Spike time codes through Morphological Learning with Binary Synapses," IEEE Transactions on Neural Networks and Learning Systems, 2015.

4. S. Roy, A. Banerjee and A. Basu, "Liquid State Machine with Dendritically Enhanced Readout for Low-Power, Neuromorphic VLSI Implementations," IEEE Transactions on Biomedical Circuits and Systems, vol. 8, pp. 681–695, Oct. 2014.

5. S. M. Islam, S. Das, S. Ghosh, S. Roy and P. N. Suganthan, "An adaptive differential evolution algorithm with novel mutation and crossover strategies for global numerical optimization," IEEE Transactions on Systems, Man, and Cybernetics – Part B, vol. 42, no. 2, pp. 482-500, 2012.

Ready to start learning?
Take This Course