What do you think?
Rate this book
Practical Recommender Systems
Practical Recommender Systems explains how recommender systems work and shows how to create and apply them for your site. After covering the basics, you’ll see how to collect user data and produce personalized recommendations. You’ll learn how to use the most popular recommendation algorithms and see examples of them in action on sites like Amazon and Netflix. Finally, the book covers scaling problems and other issues you’ll encounter as your site grows.
432 pages, ebook
First published January 1, 2019
47 people are currently reading
259 people want to read
Ratings & Reviews
Friends & Following
Create a free account to discover what your friends think of this book!
Community Reviews
Displaying 1 - 16 of 16 reviews
May 20, 2022
If you need to know, "Practical Recommender Systems" knocks it out of the park. Full stop.
Something good is happening over at Manning Publications as this is the 2nd book in short order that has just left me incredibly humbled by the amount of knowledge and effort the authors very clearly injected into their work (the other one being "Fighting Churn with Data" by Carl Gold). Kudos to Kim Falk for pouring over this text and making his vast experience designing and building recommender systems available to us all.
It is refreshing to be able to pick up a book on a subject and not run into the same tired, off-the-shelf, benchmark datasets that have been beaten to death by every data science practitioner.
In "Practical Recommender Systems" (heretofore referred to as PRS), you not only get a large, semi production-level dataset of items, users, and ratings but also a mock website that will display an end-user movie "storefront", but also all the charts, dashboards, and metrics that would be available to the back-end team of engineers and data scientists in charge of monitoring and fine-tuning the content the users see.
In turn, the "Moviegeeks" website will display the recommendation output for each user (or item) as you calculate them following each of the chapter's listings. You'll begin with non-personalized recommendations, followed by seeded recommendations, similarity functions, collaborative recs, and so on.
Trust me, dear reader, it is quite rare to find such a complete learning environment/platform to get a grasp on the material and practice. This is both a blessing and a curse, but more on that later.
Each topic is explained clearly both in terms of the logic behind each recommender algorithm and the python code for each listing. Coming from a different subfield of data science, I feel like after reading it cover-to-cover I have a good grasp of the different flavors of rec systems and their corresponding strengths/weaknesses. This is crucial for anyone looking to move on to more advanced reading.
Now, is it perfect? Nope.
The main problem with this book is that it will be somewhat "inaccessible" to those with less exposure to some of the tech the author has employed to get his code up and running in the mock environment.
Remember when I said that the "Moviegeeks" website the author will have you spin up is both a blessing and a curse? Let's start with the curse aspect: even if you carefully follow the instructions in the book's repo, chances are you will run into issues deploying the website. These are easy enough to fix (missing dependencies, hardcoded paths, incomplete dictionaries, deprecated libraries, etc...) but will discourage or outright block some people from moving forward.
Furthermore, once you get the site to load you'll want to set up access to the database that stores all the data and calculations so you can doublecheck your own work as you follow along in a local notebook. So there goes another hour...
Also, the listings that contain the code for each chapter are so convoluted and wrapped up in inherited classes and objects and function imports that it is sometimes more labor-intensive than it should be just to be able to see what's going on so that you might replicate his results (which are sometimes wrong, by the way, see the book's errata and the pull-requests in the book's repo).
Because he built the entire book's listing in the context of a commercial movie store website, he had to resort to using databases, web frameworks, SQL wrappers, etc.. to make the whole thing work. The end result is something that is uniquely effective in delivering what is ostensibly very complex material.
However, the downside is that the code seems disjointed and dependent on the underlying tech that it is used to serve it. I re-wrote much of the book's code on Jupyter notebooks and fed it the data I downloaded from the database so I could effectively separate the main idea from the execution.
The main blessing here for the more advanced users, aside from what I have mentioned before, is that there's a lot of great code that is just about production-ready. It's very "borrowable"...
I think this book could be much more successful and marketable to a larger audience if the author were to relegate the "production code" to an appendix and simplify each chapter's listing so that beginners and less tech-savvy folk may receive the enjoyment of the material and his instruction, without the added complexity of a rigid pipeline.
Even considering the aforementioned criticisms, this book deserves all stars available. Highest possible recommendation!
Something good is happening over at Manning Publications as this is the 2nd book in short order that has just left me incredibly humbled by the amount of knowledge and effort the authors very clearly injected into their work (the other one being "Fighting Churn with Data" by Carl Gold). Kudos to Kim Falk for pouring over this text and making his vast experience designing and building recommender systems available to us all.
It is refreshing to be able to pick up a book on a subject and not run into the same tired, off-the-shelf, benchmark datasets that have been beaten to death by every data science practitioner.
In "Practical Recommender Systems" (heretofore referred to as PRS), you not only get a large, semi production-level dataset of items, users, and ratings but also a mock website that will display an end-user movie "storefront", but also all the charts, dashboards, and metrics that would be available to the back-end team of engineers and data scientists in charge of monitoring and fine-tuning the content the users see.
In turn, the "Moviegeeks" website will display the recommendation output for each user (or item) as you calculate them following each of the chapter's listings. You'll begin with non-personalized recommendations, followed by seeded recommendations, similarity functions, collaborative recs, and so on.
Trust me, dear reader, it is quite rare to find such a complete learning environment/platform to get a grasp on the material and practice. This is both a blessing and a curse, but more on that later.
Each topic is explained clearly both in terms of the logic behind each recommender algorithm and the python code for each listing. Coming from a different subfield of data science, I feel like after reading it cover-to-cover I have a good grasp of the different flavors of rec systems and their corresponding strengths/weaknesses. This is crucial for anyone looking to move on to more advanced reading.
Now, is it perfect? Nope.
The main problem with this book is that it will be somewhat "inaccessible" to those with less exposure to some of the tech the author has employed to get his code up and running in the mock environment.
Remember when I said that the "Moviegeeks" website the author will have you spin up is both a blessing and a curse? Let's start with the curse aspect: even if you carefully follow the instructions in the book's repo, chances are you will run into issues deploying the website. These are easy enough to fix (missing dependencies, hardcoded paths, incomplete dictionaries, deprecated libraries, etc...) but will discourage or outright block some people from moving forward.
Furthermore, once you get the site to load you'll want to set up access to the database that stores all the data and calculations so you can doublecheck your own work as you follow along in a local notebook. So there goes another hour...
Also, the listings that contain the code for each chapter are so convoluted and wrapped up in inherited classes and objects and function imports that it is sometimes more labor-intensive than it should be just to be able to see what's going on so that you might replicate his results (which are sometimes wrong, by the way, see the book's errata and the pull-requests in the book's repo).
Because he built the entire book's listing in the context of a commercial movie store website, he had to resort to using databases, web frameworks, SQL wrappers, etc.. to make the whole thing work. The end result is something that is uniquely effective in delivering what is ostensibly very complex material.
However, the downside is that the code seems disjointed and dependent on the underlying tech that it is used to serve it. I re-wrote much of the book's code on Jupyter notebooks and fed it the data I downloaded from the database so I could effectively separate the main idea from the execution.
The main blessing here for the more advanced users, aside from what I have mentioned before, is that there's a lot of great code that is just about production-ready. It's very "borrowable"...
I think this book could be much more successful and marketable to a larger audience if the author were to relegate the "production code" to an appendix and simplify each chapter's listing so that beginners and less tech-savvy folk may receive the enjoyment of the material and his instruction, without the added complexity of a rigid pipeline.
Even considering the aforementioned criticisms, this book deserves all stars available. Highest possible recommendation!
Read
November 24, 2019I was struggling with star rating this time. To be honest, because I can't state an honest opinion on the whole book. I liked the introduction, key concepts and I was able to follow everything until 65-70% of the book, but then the entry threshold has raised quickly and myself not being an expert on statistical models and Python ... I dropped off :)
Anyway, I feel I've found what I was looking for (a primer on the topic), I liked the examples and how they were presented - in my case it was definitely NOT a time wasted.
Anyway, I feel I've found what I was looking for (a primer on the topic), I liked the examples and how they were presented - in my case it was definitely NOT a time wasted.
August 13, 2021
كان مشروع تخرجي هو بناء نظام توصية بالطعام الصحي
و عندما أقول أنظمة التوصية فهي محيطة بنا في كل مكان بدا من البسيطة منها كتوصيات القودريدز فتوصيات اليوتيوب و الفيسبوك و الأسواق الإلكترونية إلى ماهو أعقد كبناء نظام خبير يقوم الطبيب بإعطائه الأعراض و يقترح هو التشخيص بل و حتى الدواء
الحقيقة أنه تخصص العصر فعليك الإطلاع ولو بشكل سطحي خاصة و أن هذا المجال يتطور بشكل رهيب في ظروف زمنية قصيرة جدا
لم أرد أن أطل كثيرا بإمكاني الإجابة عن الأسئلة إن وجدت
بعد أن بحثت على الشبكة ، أنا مقتنع بأن هذا هو أفضل كتاب يقدم لك المجال بسلاسة
لن أقول إنني قرأت الكتاب بأكمله ، لكنه كان مفيدًا حقًا بالنسبة لي.
في النهاية أعتقد أن ترجمة الكتاب إلى اللغة العربية فكرة جيدة خاصة بسبب النقص الشديد المعلومات حول هذا المجال باللغة العربية.
My end of study project was to build a healthy food recommender system
and after I searched on the web I have convinced that this is the best book that will introduce the domain to you smoothly
I will not say that I've read the whole book, but it was really useful to me.
at the end I think that to translate the book to Arabic is a good idea especially because of the lack of information about this field in Arabic.
و عندما أقول أنظمة التوصية فهي محيطة بنا في كل مكان بدا من البسيطة منها كتوصيات القودريدز فتوصيات اليوتيوب و الفيسبوك و الأسواق الإلكترونية إلى ماهو أعقد كبناء نظام خبير يقوم الطبيب بإعطائه الأعراض و يقترح هو التشخيص بل و حتى الدواء
الحقيقة أنه تخصص العصر فعليك الإطلاع ولو بشكل سطحي خاصة و أن هذا المجال يتطور بشكل رهيب في ظروف زمنية قصيرة جدا
لم أرد أن أطل كثيرا بإمكاني الإجابة عن الأسئلة إن وجدت
بعد أن بحثت على الشبكة ، أنا مقتنع بأن هذا هو أفضل كتاب يقدم لك المجال بسلاسة
لن أقول إنني قرأت الكتاب بأكمله ، لكنه كان مفيدًا حقًا بالنسبة لي.
في النهاية أعتقد أن ترجمة الكتاب إلى اللغة العربية فكرة جيدة خاصة بسبب النقص الشديد المعلومات حول هذا المجال باللغة العربية.
My end of study project was to build a healthy food recommender system
and after I searched on the web I have convinced that this is the best book that will introduce the domain to you smoothly
I will not say that I've read the whole book, but it was really useful to me.
at the end I think that to translate the book to Arabic is a good idea especially because of the lack of information about this field in Arabic.
December 17, 2021
Excellent practical intro to recsys. What makes it so valuable is the app in the book and the blend of theory plus real code, which is hard to find since most of recsys is still academic papers at this point.
February 11, 2019
Very practical book about recommendation systems.
April 9, 2019
Excellent book on how to implement recommendation systems! This book has been very helpful in my search ranking and recommendation projects at work, I found the chapters on matrix factorization and learning-to-rank to be especially useful. The material strikes the right balance between theory and practical implementation. The material is easy to understand and has a lot of charts and diagrams to support the written text. I would definitely recommend this book to anyone who is interested in learning about recommendation systems!
November 15, 2021
Pretty good book for refreshing some of the ideas in classical recommender system. I realized that i am already familiar with most of the materials in the book such as BPR, SVD, latent matrix factorization (funk SVD?), association rule mining, similarity metrics such as pearson, jaccard, cosine. But the book is a good refreshment and the math is actually quite light. The only thing that i feel missing is that the book is missing other important recommender algorithms from deep learning such as neural matrix factorization, factorization machine, wide and deep learning. The implementation of BPR can be significantly simpler and easier to grasp and implement if the author brings deep learning framework such as keras or pytorch.
July 29, 2019
I review the book until Chapter 8, and find that it is pretty hard to continue, why? Because the Django
demo code in my local machine cannot run somehow, I got pretty confused. That's the only drawback I find so far. All in all, the book is good, I get some new ideas how I can build a recommender system. I would recommend this book to someone who has some experience on the recommender system already!
demo code in my local machine cannot run somehow, I got pretty confused. That's the only drawback I find so far. All in all, the book is good, I get some new ideas how I can build a recommender system. I would recommend this book to someone who has some experience on the recommender system already!
June 14, 2021
Great overview of all the recommenders possible, also super-nice to have some practical pieces of code and to combine developer and data scientist's approaches. Easy to read. Some theoretical parts were hard to comprehend but I guess it is just the subject itself which is hard. Hopefully I'll use some hints from the book for my current work project.
July 5, 2023
A great book to get you started in recommendation algorithms. This is by no means a comprehensive text but it will give you a good high-level understanding of different aspects of any recommender system.
January 25, 2024
Actually used this for work! This book does a great job at providing broader context as well as a good level of detail. Goes into not just what a recommender system should do but also what it should not do.
January 6, 2021
the best book for recommendation systems
January 25, 2021
A comprehensive reading on recommender systems and one that is quite recent. Examples codebase, MovieGeek is good and fun to read.
March 28, 2021
Great book that helps me implement my own rec system even without machine learning.
August 25, 2023
Too loose
March 12, 2025
A book introduction but doesn’t cover many modern approaches
Displaying 1 - 16 of 16 reviews