SlideShare a Scribd company logo
Building a modern recommender system
Alex Konduforov
Data Science Competence Leader
AltexSoft
Intro to RecSys
RecSys are everywhere
Evolution
Tapestry (Xerox Lab) - emails
1992
GroupLens project (UoM) - UseNet,
MovieLens
1994
Amazon - item-based collaborative
filtering
1998
Netflix - Cinematch, Netflix Prize
2006
Spotify - uses CNNs for audio analysis
2013
Youtube
2010 - naïve recommender
2016 - Deep Neural Nets by DeepMind
2016
Types of recommendations
Product recommendations Content recommendations
• Choice overload – infinite "shelves"
• Surprising suggestions
• Higher conversion, more purchases
• Preserving customer attention
• Competitive advantage across industry
Why do we need them?
Main approaches
Recommender systems
Content-based filtering Collaborative filtering Hybrid systems
Memory-based
(neighborhood-based)
Model-based
(mix of content and collaborative
methods)
Content-based methods
Uses similarity between items to recommend items similar to what the user likes.
2 approaches:
1) Build vector for item
2) Find similar item vectors
(cosine, dot product, Euclidian, etc.)
1) Build vectors for user items
2) Derive a user vector from them
3) Find similar item vectors
Content-based filtering
Image source
User-centered modeling
Item-centered modeling
Content-based modeling
Image source
Collaborative-based methods
Uses interactions between users and items simultaneously to provide
recommendations.
Collaborative filtering
Image source
User-user CF
Neighborhood-based CF
Image source
Item-item CF
Neighborhood-based CF
Image source
User-user:
• Users have not many interactions
• KNN is sensitive to single interactions (high
variance)
• More personalized results (low bias)
Neighborhood-based CF
Item-item:
• Items have many users interacted with them
• KNN is less sensitive to single interactions (lower
variance)
• Less personalized results (higher bias)
• Works better for new users (not enough history)
• More likely to converge
General issues:
• KNN is time consuming, doesn't scale well
• Higher effect of "rich-get-richer" for popular items
Also referred as Latent factor methods
Approaches:
1. SVD
2. Matrix factorization
3. Neural Networks
Model-based CF
Model-based CF: Matrix Factorization
Image source
Matrix Factorization
Training:
• Initialize P and Q matrices with small random numbers
• Teach P and Q
• Alternating Least Squares
• Stochastic Gradient Descent
Predictions:
MF algorithm
MF example
Latent features are calculated via MF:
User-item score is the dot product:
Item-item similarity is the cosine similarity:
• MF with Biases: handling bias of some users giving higher ratings than others
• MF with Side Features: adding data to handle the cold start problem (i.e. user occupation)
• MF with Temporal Features: handling temporal changes of the data (i.e. occupation change)
• Factorization Machine: extra item features + higher order interactions
• MF with Fixture of tastes: give users several tastes
• Variational MF
MF improvements
Source article
Why?
• Collaborative filtering uses robust approach based on similarities between customer tastes
and can make cross-genre(category) recommendations
• Pure CF lacks information about items and suffer from sparsity in data
• Content-based filtering may find deep similarities between items and suggest novel and
surprising new items
• Only CBF can bring value from images, sounds, texts
• All modern implementations are hybrid
Hybrid
More info: https://www.researchgate.net/publication/263377228_Hybrid_Recommender_Systems_Survey_and_Experiments
Deep Learning for RecSys
Deep Learning revolution
• Modeling the non-linear interactions in the data
• Feature extraction directly from the content:
• Image, audio, text
• Ability to use heterogeneous data (interactions, content) in the same model
• Powerful for sequential modeling tasks (next item prediction, session-based recommendations)
• Better representation learning of users and items for CF
Deep Learning for RecSys
Bringing NNs to CF
Neural Collaborative Filtering Deep Factorization Machine
Several more NN-based improvements can be found here: https://towardsdatascience.com/recsys-series-part-5-neural-
matrix-factorization-for-collaborative-filtering-a0aebfe15883
• Prior to Deep Learning approach used Matrix Factorization
• 2 Neural Networks:
• Candidates generation: broad personalization via CF
(retrieves only couple of hundreds of videos)
• Ranking: assigns a score to each video using a rich
set of features from item and user
• Uses implicit signals (full watching is positive), not explicit
(thumbs up/down)
• Use video "age" as a feature to recommend more new
content
• Rely mostly on A/B testing results
https://research.google/pubs/pub45530/
Youtube recommendations
• Uses implicit customer feedback
• Back in 2014 used Weighted MF
• Used MFCC and CNNs to extract features from music –
helps with cold start and finding unpopular tracks
• Used NLP to extract features from song texts and other
textual information
• Basically uses a hybrid recommender system
https://benanne.github.io/2014/08/05/spotify-cnns.html
https://www.oreilly.com/radar/personalization-of-spotify-home-
and-tensorflow/
Spotify recommendations
• Autoencoder-based recommendations:
• to learn the lower-dimensional feature representations at the bottleneck layer
• to fill in the blanks of the user-item interaction matrix directly in the reconstruction layer
• CNN-based recommendations:
• to extract features from images
• to extract features from audio and video
• to extract features from texts
• RNN-based recommendations
• to extract sequential patterns in session-based tasks
• and many others: https://towardsdatascience.com/recommendation-system-series-part-2-the-10-
categories-of-deep-recommendation-systems-that-189d60287b58
Other DL approaches
Metrics:
• When predicting rating: MSE, RMSE, etc.
• When predicting binary output: Accuracy, Precision, Recall, F1
• When predicting top N: MAP@N, MAR@N
• Coverage: % of items which are recommended (average CF predicts ~8-10%)
Real-world testing:
• CTR or CR (product, ads, etc.)
• Avg Time spent or Customer retention (content)
• A/B testing
• Serendipity: pleasant surprise, unintended discovery
Evaluation
• Novelty
• Diversity
• Serendipity
• Interpretability
• Adaptation
Qualities of a good recommender
• https://towardsdatascience.com/introduction-to-recommender-systems-6c66cf15ada
• https://towardsdatascience.com/recommendation-system-series-part-1-an-executive-guide-to-
building-recommendation-system-608f83e2630a
• https://medium.com/libreai/a-glimpse-into-deep-learning-for-recommender-systems-
d66ae0681775
• ADD: https://towardsdatascience.com/modern-recommender-systems-a0c727609aa8
Materials
Alex Konduforov
alexander.konduforov@gmail.com
@alex_konduforov
Thank you!

More Related Content

More from Lviv Startup Club

Khristina Pototska: Steering the Ship: Product Management in Startups vs. Glo...
Khristina Pototska: Steering the Ship: Product Management in Startups vs. Glo...Khristina Pototska: Steering the Ship: Product Management in Startups vs. Glo...
Khristina Pototska: Steering the Ship: Product Management in Startups vs. Glo...
Lviv Startup Club
 
Oleksandr Buratynskyi: Як Agile Coach мікроменеджером став 🙃 (UA)
Oleksandr Buratynskyi: Як Agile Coach мікроменеджером став 🙃 (UA)Oleksandr Buratynskyi: Як Agile Coach мікроменеджером став 🙃 (UA)
Oleksandr Buratynskyi: Як Agile Coach мікроменеджером став 🙃 (UA)
Lviv Startup Club
 
Igor Protsenko: Difference between outsourcing and product companies for prod...
Igor Protsenko: Difference between outsourcing and product companies for prod...Igor Protsenko: Difference between outsourcing and product companies for prod...
Igor Protsenko: Difference between outsourcing and product companies for prod...
Lviv Startup Club
 
Kseniya Leshchenko: Shared development support service model as the way to ma...
Kseniya Leshchenko: Shared development support service model as the way to ma...Kseniya Leshchenko: Shared development support service model as the way to ma...
Kseniya Leshchenko: Shared development support service model as the way to ma...
Lviv Startup Club
 
Valeriy Kozlov: Taming the Startup Chaos: GTD for Founders & Small Teams (UA)
Valeriy Kozlov: Taming the Startup Chaos: GTD for Founders & Small Teams (UA)Valeriy Kozlov: Taming the Startup Chaos: GTD for Founders & Small Teams (UA)
Valeriy Kozlov: Taming the Startup Chaos: GTD for Founders & Small Teams (UA)
Lviv Startup Club
 
Anna Kompanets: Проблеми впровадження проєктів, про які б ви ніколи не подума...
Anna Kompanets: Проблеми впровадження проєктів, про які б ви ніколи не подума...Anna Kompanets: Проблеми впровадження проєктів, про які б ви ніколи не подума...
Anna Kompanets: Проблеми впровадження проєктів, про які б ви ніколи не подума...
Lviv Startup Club
 
Viktoriia Honcharova: PMI: нова стратегія розвитку управління проєктами (UA)
Viktoriia Honcharova: PMI: нова стратегія розвитку управління проєктами (UA)Viktoriia Honcharova: PMI: нова стратегія розвитку управління проєктами (UA)
Viktoriia Honcharova: PMI: нова стратегія розвитку управління проєктами (UA)
Lviv Startup Club
 
Andrii Mandrika: Як системно допомагати ЗСУ, використовуючи продуктовий підхі...
Andrii Mandrika: Як системно допомагати ЗСУ, використовуючи продуктовий підхі...Andrii Mandrika: Як системно допомагати ЗСУ, використовуючи продуктовий підхі...
Andrii Mandrika: Як системно допомагати ЗСУ, використовуючи продуктовий підхі...
Lviv Startup Club
 
Michael Vidyakin: From Vision to Victory: Mastering the Project-Strategy Conn...
Michael Vidyakin: From Vision to Victory: Mastering the Project-Strategy Conn...Michael Vidyakin: From Vision to Victory: Mastering the Project-Strategy Conn...
Michael Vidyakin: From Vision to Victory: Mastering the Project-Strategy Conn...
Lviv Startup Club
 
Kateryna Kubasova: Абстрактне Оксфордське лідерство конкретному українському ...
Kateryna Kubasova: Абстрактне Оксфордське лідерство конкретному українському ...Kateryna Kubasova: Абстрактне Оксфордське лідерство конкретному українському ...
Kateryna Kubasova: Абстрактне Оксфордське лідерство конкретному українському ...
Lviv Startup Club
 
Andrii Salii: Навіщо публічному сектору NPS: будуємо довіру через відкритість...
Andrii Salii: Навіщо публічному сектору NPS: будуємо довіру через відкритість...Andrii Salii: Навіщо публічному сектору NPS: будуємо довіру через відкритість...
Andrii Salii: Навіщо публічному сектору NPS: будуємо довіру через відкритість...
Lviv Startup Club
 
Anton Hlazkov: Впровадження змін – це процес чи проєкт? Чому важливо розуміти...
Anton Hlazkov: Впровадження змін – це процес чи проєкт? Чому важливо розуміти...Anton Hlazkov: Впровадження змін – це процес чи проєкт? Чому важливо розуміти...
Anton Hlazkov: Впровадження змін – це процес чи проєкт? Чому важливо розуміти...
Lviv Startup Club
 
Yana Bort: Ритм організації. Чи можливо синхронізувати великий ентерпрайз за ...
Yana Bort: Ритм організації. Чи можливо синхронізувати великий ентерпрайз за ...Yana Bort: Ритм організації. Чи можливо синхронізувати великий ентерпрайз за ...
Yana Bort: Ритм організації. Чи можливо синхронізувати великий ентерпрайз за ...
Lviv Startup Club
 
Nikita Artemchuk: Навчання та розвиток продакт менеджера (UA)
Nikita Artemchuk: Навчання та розвиток продакт менеджера (UA)Nikita Artemchuk: Навчання та розвиток продакт менеджера (UA)
Nikita Artemchuk: Навчання та розвиток продакт менеджера (UA)
Lviv Startup Club
 
Mykyta Melnyk: Досвід провадження AI Driven Development, кейси використання т...
Mykyta Melnyk: Досвід провадження AI Driven Development, кейси використання т...Mykyta Melnyk: Досвід провадження AI Driven Development, кейси використання т...
Mykyta Melnyk: Досвід провадження AI Driven Development, кейси використання т...
Lviv Startup Club
 
Oleksandr Marchenko: Складності росту продуктових команд у сучасних умовах (UA)
Oleksandr Marchenko: Складності росту продуктових команд у сучасних умовах (UA)Oleksandr Marchenko: Складності росту продуктових команд у сучасних умовах (UA)
Oleksandr Marchenko: Складності росту продуктових команд у сучасних умовах (UA)
Lviv Startup Club
 
Veronica Rodionova: Подолання опору впровадженню Agile процесів у командах (UA)
Veronica Rodionova: Подолання опору впровадженню Agile процесів у командах (UA)Veronica Rodionova: Подолання опору впровадженню Agile процесів у командах (UA)
Veronica Rodionova: Подолання опору впровадженню Agile процесів у командах (UA)
Lviv Startup Club
 
Iryna Koberniuk: Implementing Major Changes: How to Effectively Update a Prod...
Iryna Koberniuk: Implementing Major Changes: How to Effectively Update a Prod...Iryna Koberniuk: Implementing Major Changes: How to Effectively Update a Prod...
Iryna Koberniuk: Implementing Major Changes: How to Effectively Update a Prod...
Lviv Startup Club
 
Hanna Klimushka: Прокачка продуктового мислення для проєктного менеджера (UA)
Hanna Klimushka: Прокачка продуктового мислення для проєктного менеджера (UA)Hanna Klimushka: Прокачка продуктового мислення для проєктного менеджера (UA)
Hanna Klimushka: Прокачка продуктового мислення для проєктного менеджера (UA)
Lviv Startup Club
 
Ihor Pavlenko: PMO Risk Management (UA).
Ihor Pavlenko: PMO Risk Management (UA).Ihor Pavlenko: PMO Risk Management (UA).
Ihor Pavlenko: PMO Risk Management (UA).
Lviv Startup Club
 

More from Lviv Startup Club (20)

Khristina Pototska: Steering the Ship: Product Management in Startups vs. Glo...
Khristina Pototska: Steering the Ship: Product Management in Startups vs. Glo...Khristina Pototska: Steering the Ship: Product Management in Startups vs. Glo...
Khristina Pototska: Steering the Ship: Product Management in Startups vs. Glo...
 
Oleksandr Buratynskyi: Як Agile Coach мікроменеджером став 🙃 (UA)
Oleksandr Buratynskyi: Як Agile Coach мікроменеджером став 🙃 (UA)Oleksandr Buratynskyi: Як Agile Coach мікроменеджером став 🙃 (UA)
Oleksandr Buratynskyi: Як Agile Coach мікроменеджером став 🙃 (UA)
 
Igor Protsenko: Difference between outsourcing and product companies for prod...
Igor Protsenko: Difference between outsourcing and product companies for prod...Igor Protsenko: Difference between outsourcing and product companies for prod...
Igor Protsenko: Difference between outsourcing and product companies for prod...
 
Kseniya Leshchenko: Shared development support service model as the way to ma...
Kseniya Leshchenko: Shared development support service model as the way to ma...Kseniya Leshchenko: Shared development support service model as the way to ma...
Kseniya Leshchenko: Shared development support service model as the way to ma...
 
Valeriy Kozlov: Taming the Startup Chaos: GTD for Founders & Small Teams (UA)
Valeriy Kozlov: Taming the Startup Chaos: GTD for Founders & Small Teams (UA)Valeriy Kozlov: Taming the Startup Chaos: GTD for Founders & Small Teams (UA)
Valeriy Kozlov: Taming the Startup Chaos: GTD for Founders & Small Teams (UA)
 
Anna Kompanets: Проблеми впровадження проєктів, про які б ви ніколи не подума...
Anna Kompanets: Проблеми впровадження проєктів, про які б ви ніколи не подума...Anna Kompanets: Проблеми впровадження проєктів, про які б ви ніколи не подума...
Anna Kompanets: Проблеми впровадження проєктів, про які б ви ніколи не подума...
 
Viktoriia Honcharova: PMI: нова стратегія розвитку управління проєктами (UA)
Viktoriia Honcharova: PMI: нова стратегія розвитку управління проєктами (UA)Viktoriia Honcharova: PMI: нова стратегія розвитку управління проєктами (UA)
Viktoriia Honcharova: PMI: нова стратегія розвитку управління проєктами (UA)
 
Andrii Mandrika: Як системно допомагати ЗСУ, використовуючи продуктовий підхі...
Andrii Mandrika: Як системно допомагати ЗСУ, використовуючи продуктовий підхі...Andrii Mandrika: Як системно допомагати ЗСУ, використовуючи продуктовий підхі...
Andrii Mandrika: Як системно допомагати ЗСУ, використовуючи продуктовий підхі...
 
Michael Vidyakin: From Vision to Victory: Mastering the Project-Strategy Conn...
Michael Vidyakin: From Vision to Victory: Mastering the Project-Strategy Conn...Michael Vidyakin: From Vision to Victory: Mastering the Project-Strategy Conn...
Michael Vidyakin: From Vision to Victory: Mastering the Project-Strategy Conn...
 
Kateryna Kubasova: Абстрактне Оксфордське лідерство конкретному українському ...
Kateryna Kubasova: Абстрактне Оксфордське лідерство конкретному українському ...Kateryna Kubasova: Абстрактне Оксфордське лідерство конкретному українському ...
Kateryna Kubasova: Абстрактне Оксфордське лідерство конкретному українському ...
 
Andrii Salii: Навіщо публічному сектору NPS: будуємо довіру через відкритість...
Andrii Salii: Навіщо публічному сектору NPS: будуємо довіру через відкритість...Andrii Salii: Навіщо публічному сектору NPS: будуємо довіру через відкритість...
Andrii Salii: Навіщо публічному сектору NPS: будуємо довіру через відкритість...
 
Anton Hlazkov: Впровадження змін – це процес чи проєкт? Чому важливо розуміти...
Anton Hlazkov: Впровадження змін – це процес чи проєкт? Чому важливо розуміти...Anton Hlazkov: Впровадження змін – це процес чи проєкт? Чому важливо розуміти...
Anton Hlazkov: Впровадження змін – це процес чи проєкт? Чому важливо розуміти...
 
Yana Bort: Ритм організації. Чи можливо синхронізувати великий ентерпрайз за ...
Yana Bort: Ритм організації. Чи можливо синхронізувати великий ентерпрайз за ...Yana Bort: Ритм організації. Чи можливо синхронізувати великий ентерпрайз за ...
Yana Bort: Ритм організації. Чи можливо синхронізувати великий ентерпрайз за ...
 
Nikita Artemchuk: Навчання та розвиток продакт менеджера (UA)
Nikita Artemchuk: Навчання та розвиток продакт менеджера (UA)Nikita Artemchuk: Навчання та розвиток продакт менеджера (UA)
Nikita Artemchuk: Навчання та розвиток продакт менеджера (UA)
 
Mykyta Melnyk: Досвід провадження AI Driven Development, кейси використання т...
Mykyta Melnyk: Досвід провадження AI Driven Development, кейси використання т...Mykyta Melnyk: Досвід провадження AI Driven Development, кейси використання т...
Mykyta Melnyk: Досвід провадження AI Driven Development, кейси використання т...
 
Oleksandr Marchenko: Складності росту продуктових команд у сучасних умовах (UA)
Oleksandr Marchenko: Складності росту продуктових команд у сучасних умовах (UA)Oleksandr Marchenko: Складності росту продуктових команд у сучасних умовах (UA)
Oleksandr Marchenko: Складності росту продуктових команд у сучасних умовах (UA)
 
Veronica Rodionova: Подолання опору впровадженню Agile процесів у командах (UA)
Veronica Rodionova: Подолання опору впровадженню Agile процесів у командах (UA)Veronica Rodionova: Подолання опору впровадженню Agile процесів у командах (UA)
Veronica Rodionova: Подолання опору впровадженню Agile процесів у командах (UA)
 
Iryna Koberniuk: Implementing Major Changes: How to Effectively Update a Prod...
Iryna Koberniuk: Implementing Major Changes: How to Effectively Update a Prod...Iryna Koberniuk: Implementing Major Changes: How to Effectively Update a Prod...
Iryna Koberniuk: Implementing Major Changes: How to Effectively Update a Prod...
 
Hanna Klimushka: Прокачка продуктового мислення для проєктного менеджера (UA)
Hanna Klimushka: Прокачка продуктового мислення для проєктного менеджера (UA)Hanna Klimushka: Прокачка продуктового мислення для проєктного менеджера (UA)
Hanna Klimushka: Прокачка продуктового мислення для проєктного менеджера (UA)
 
Ihor Pavlenko: PMO Risk Management (UA).
Ihor Pavlenko: PMO Risk Management (UA).Ihor Pavlenko: PMO Risk Management (UA).
Ihor Pavlenko: PMO Risk Management (UA).
 

Recently uploaded

Business storytelling: key ingredients to a story
Business storytelling: key ingredients to a storyBusiness storytelling: key ingredients to a story
Business storytelling: key ingredients to a story
Alexandra Fulford
 
Digital Marketing with a Focus on Sustainability
Digital Marketing with a Focus on SustainabilityDigital Marketing with a Focus on Sustainability
Digital Marketing with a Focus on Sustainability
sssourabhsharma
 
How are Lilac French Bulldogs Beauty Charming the World and Capturing Hearts....
How are Lilac French Bulldogs Beauty Charming the World and Capturing Hearts....How are Lilac French Bulldogs Beauty Charming the World and Capturing Hearts....
How are Lilac French Bulldogs Beauty Charming the World and Capturing Hearts....
Lacey Max
 
Registered-Establishment-List-in-Uttarakhand-pdf.pdf
Registered-Establishment-List-in-Uttarakhand-pdf.pdfRegistered-Establishment-List-in-Uttarakhand-pdf.pdf
Registered-Establishment-List-in-Uttarakhand-pdf.pdf
dazzjoker
 
GKohler - Retail Scavenger Hunt Presentation
GKohler - Retail Scavenger Hunt PresentationGKohler - Retail Scavenger Hunt Presentation
GKohler - Retail Scavenger Hunt Presentation
GraceKohler1
 
Best Forex Brokers Comparison in INDIA 2024
Best Forex Brokers Comparison in INDIA 2024Best Forex Brokers Comparison in INDIA 2024
Best Forex Brokers Comparison in INDIA 2024
Top Forex Brokers Review
 
HOW TO START UP A COMPANY A STEP-BY-STEP GUIDE.pdf
HOW TO START UP A COMPANY A STEP-BY-STEP GUIDE.pdfHOW TO START UP A COMPANY A STEP-BY-STEP GUIDE.pdf
HOW TO START UP A COMPANY A STEP-BY-STEP GUIDE.pdf
46adnanshahzad
 
一比一原版(QMUE毕业证书)英国爱丁堡玛格丽特女王大学毕业证文凭如何办理
一比一原版(QMUE毕业证书)英国爱丁堡玛格丽特女王大学毕业证文凭如何办理一比一原版(QMUE毕业证书)英国爱丁堡玛格丽特女王大学毕业证文凭如何办理
一比一原版(QMUE毕业证书)英国爱丁堡玛格丽特女王大学毕业证文凭如何办理
taqyea
 
The latest Heat Pump Manual from Newentide
The latest Heat Pump Manual from NewentideThe latest Heat Pump Manual from Newentide
The latest Heat Pump Manual from Newentide
JoeYangGreatMachiner
 
Call8328958814 satta matka Kalyan result satta guessing
Call8328958814 satta matka Kalyan result satta guessingCall8328958814 satta matka Kalyan result satta guessing
Call8328958814 satta matka Kalyan result satta guessing
➑➌➋➑➒➎➑➑➊➍
 
The Most Inspiring Entrepreneurs to Follow in 2024.pdf
The Most Inspiring Entrepreneurs to Follow in 2024.pdfThe Most Inspiring Entrepreneurs to Follow in 2024.pdf
The Most Inspiring Entrepreneurs to Follow in 2024.pdf
thesiliconleaders
 
Dpboss Matka Guessing Satta Matta Matka Kalyan Chart Satta Matka
Dpboss Matka Guessing Satta Matta Matka Kalyan Chart Satta MatkaDpboss Matka Guessing Satta Matta Matka Kalyan Chart Satta Matka
Dpboss Matka Guessing Satta Matta Matka Kalyan Chart Satta Matka
➒➌➎➏➑➐➋➑➐➐Dpboss Matka Guessing Satta Matka Kalyan Chart Indian Matka
 
Lundin Gold Corporate Presentation - June 2024
Lundin Gold Corporate Presentation - June 2024Lundin Gold Corporate Presentation - June 2024
Lundin Gold Corporate Presentation - June 2024
Adnet Communications
 
How MJ Global Leads the Packaging Industry.pdf
How MJ Global Leads the Packaging Industry.pdfHow MJ Global Leads the Packaging Industry.pdf
How MJ Global Leads the Packaging Industry.pdf
MJ Global
 
Industrial Tech SW: Category Renewal and Creation
Industrial Tech SW:  Category Renewal and CreationIndustrial Tech SW:  Category Renewal and Creation
Industrial Tech SW: Category Renewal and Creation
Christian Dahlen
 
Best practices for project execution and delivery
Best practices for project execution and deliveryBest practices for project execution and delivery
Best practices for project execution and delivery
CLIVE MINCHIN
 
Digital Transformation Frameworks: Driving Digital Excellence
Digital Transformation Frameworks: Driving Digital ExcellenceDigital Transformation Frameworks: Driving Digital Excellence
Digital Transformation Frameworks: Driving Digital Excellence
Operational Excellence Consulting
 
Top mailing list providers in the USA.pptx
Top mailing list providers in the USA.pptxTop mailing list providers in the USA.pptx
Top mailing list providers in the USA.pptx
JeremyPeirce1
 
Taurus Zodiac Sign: Unveiling the Traits, Dates, and Horoscope Insights of th...
Taurus Zodiac Sign: Unveiling the Traits, Dates, and Horoscope Insights of th...Taurus Zodiac Sign: Unveiling the Traits, Dates, and Horoscope Insights of th...
Taurus Zodiac Sign: Unveiling the Traits, Dates, and Horoscope Insights of th...
my Pandit
 
❼❷⓿❺❻❷❽❷❼❽ Dpboss Matka Result Satta Matka Guessing Satta Fix jodi Kalyan Fin...
❼❷⓿❺❻❷❽❷❼❽ Dpboss Matka Result Satta Matka Guessing Satta Fix jodi Kalyan Fin...❼❷⓿❺❻❷❽❷❼❽ Dpboss Matka Result Satta Matka Guessing Satta Fix jodi Kalyan Fin...
❼❷⓿❺❻❷❽❷❼❽ Dpboss Matka Result Satta Matka Guessing Satta Fix jodi Kalyan Fin...
❼❷⓿❺❻❷❽❷❼❽ Dpboss Kalyan Satta Matka Guessing Matka Result Main Bazar chart
 

Recently uploaded (20)

Business storytelling: key ingredients to a story
Business storytelling: key ingredients to a storyBusiness storytelling: key ingredients to a story
Business storytelling: key ingredients to a story
 
Digital Marketing with a Focus on Sustainability
Digital Marketing with a Focus on SustainabilityDigital Marketing with a Focus on Sustainability
Digital Marketing with a Focus on Sustainability
 
How are Lilac French Bulldogs Beauty Charming the World and Capturing Hearts....
How are Lilac French Bulldogs Beauty Charming the World and Capturing Hearts....How are Lilac French Bulldogs Beauty Charming the World and Capturing Hearts....
How are Lilac French Bulldogs Beauty Charming the World and Capturing Hearts....
 
Registered-Establishment-List-in-Uttarakhand-pdf.pdf
Registered-Establishment-List-in-Uttarakhand-pdf.pdfRegistered-Establishment-List-in-Uttarakhand-pdf.pdf
Registered-Establishment-List-in-Uttarakhand-pdf.pdf
 
GKohler - Retail Scavenger Hunt Presentation
GKohler - Retail Scavenger Hunt PresentationGKohler - Retail Scavenger Hunt Presentation
GKohler - Retail Scavenger Hunt Presentation
 
Best Forex Brokers Comparison in INDIA 2024
Best Forex Brokers Comparison in INDIA 2024Best Forex Brokers Comparison in INDIA 2024
Best Forex Brokers Comparison in INDIA 2024
 
HOW TO START UP A COMPANY A STEP-BY-STEP GUIDE.pdf
HOW TO START UP A COMPANY A STEP-BY-STEP GUIDE.pdfHOW TO START UP A COMPANY A STEP-BY-STEP GUIDE.pdf
HOW TO START UP A COMPANY A STEP-BY-STEP GUIDE.pdf
 
一比一原版(QMUE毕业证书)英国爱丁堡玛格丽特女王大学毕业证文凭如何办理
一比一原版(QMUE毕业证书)英国爱丁堡玛格丽特女王大学毕业证文凭如何办理一比一原版(QMUE毕业证书)英国爱丁堡玛格丽特女王大学毕业证文凭如何办理
一比一原版(QMUE毕业证书)英国爱丁堡玛格丽特女王大学毕业证文凭如何办理
 
The latest Heat Pump Manual from Newentide
The latest Heat Pump Manual from NewentideThe latest Heat Pump Manual from Newentide
The latest Heat Pump Manual from Newentide
 
Call8328958814 satta matka Kalyan result satta guessing
Call8328958814 satta matka Kalyan result satta guessingCall8328958814 satta matka Kalyan result satta guessing
Call8328958814 satta matka Kalyan result satta guessing
 
The Most Inspiring Entrepreneurs to Follow in 2024.pdf
The Most Inspiring Entrepreneurs to Follow in 2024.pdfThe Most Inspiring Entrepreneurs to Follow in 2024.pdf
The Most Inspiring Entrepreneurs to Follow in 2024.pdf
 
Dpboss Matka Guessing Satta Matta Matka Kalyan Chart Satta Matka
Dpboss Matka Guessing Satta Matta Matka Kalyan Chart Satta MatkaDpboss Matka Guessing Satta Matta Matka Kalyan Chart Satta Matka
Dpboss Matka Guessing Satta Matta Matka Kalyan Chart Satta Matka
 
Lundin Gold Corporate Presentation - June 2024
Lundin Gold Corporate Presentation - June 2024Lundin Gold Corporate Presentation - June 2024
Lundin Gold Corporate Presentation - June 2024
 
How MJ Global Leads the Packaging Industry.pdf
How MJ Global Leads the Packaging Industry.pdfHow MJ Global Leads the Packaging Industry.pdf
How MJ Global Leads the Packaging Industry.pdf
 
Industrial Tech SW: Category Renewal and Creation
Industrial Tech SW:  Category Renewal and CreationIndustrial Tech SW:  Category Renewal and Creation
Industrial Tech SW: Category Renewal and Creation
 
Best practices for project execution and delivery
Best practices for project execution and deliveryBest practices for project execution and delivery
Best practices for project execution and delivery
 
Digital Transformation Frameworks: Driving Digital Excellence
Digital Transformation Frameworks: Driving Digital ExcellenceDigital Transformation Frameworks: Driving Digital Excellence
Digital Transformation Frameworks: Driving Digital Excellence
 
Top mailing list providers in the USA.pptx
Top mailing list providers in the USA.pptxTop mailing list providers in the USA.pptx
Top mailing list providers in the USA.pptx
 
Taurus Zodiac Sign: Unveiling the Traits, Dates, and Horoscope Insights of th...
Taurus Zodiac Sign: Unveiling the Traits, Dates, and Horoscope Insights of th...Taurus Zodiac Sign: Unveiling the Traits, Dates, and Horoscope Insights of th...
Taurus Zodiac Sign: Unveiling the Traits, Dates, and Horoscope Insights of th...
 
❼❷⓿❺❻❷❽❷❼❽ Dpboss Matka Result Satta Matka Guessing Satta Fix jodi Kalyan Fin...
❼❷⓿❺❻❷❽❷❼❽ Dpboss Matka Result Satta Matka Guessing Satta Fix jodi Kalyan Fin...❼❷⓿❺❻❷❽❷❼❽ Dpboss Matka Result Satta Matka Guessing Satta Fix jodi Kalyan Fin...
❼❷⓿❺❻❷❽❷❼❽ Dpboss Matka Result Satta Matka Guessing Satta Fix jodi Kalyan Fin...
 

Alexander Konduforov: Building modern recommender systems

  • 1. Building a modern recommender system Alex Konduforov Data Science Competence Leader AltexSoft
  • 4. Evolution Tapestry (Xerox Lab) - emails 1992 GroupLens project (UoM) - UseNet, MovieLens 1994 Amazon - item-based collaborative filtering 1998 Netflix - Cinematch, Netflix Prize 2006 Spotify - uses CNNs for audio analysis 2013 Youtube 2010 - naïve recommender 2016 - Deep Neural Nets by DeepMind 2016
  • 5. Types of recommendations Product recommendations Content recommendations
  • 6. • Choice overload – infinite "shelves" • Surprising suggestions • Higher conversion, more purchases • Preserving customer attention • Competitive advantage across industry Why do we need them?
  • 7. Main approaches Recommender systems Content-based filtering Collaborative filtering Hybrid systems Memory-based (neighborhood-based) Model-based (mix of content and collaborative methods)
  • 9. Uses similarity between items to recommend items similar to what the user likes. 2 approaches: 1) Build vector for item 2) Find similar item vectors (cosine, dot product, Euclidian, etc.) 1) Build vectors for user items 2) Derive a user vector from them 3) Find similar item vectors Content-based filtering Image source
  • 12. Uses interactions between users and items simultaneously to provide recommendations. Collaborative filtering Image source
  • 15. User-user: • Users have not many interactions • KNN is sensitive to single interactions (high variance) • More personalized results (low bias) Neighborhood-based CF Item-item: • Items have many users interacted with them • KNN is less sensitive to single interactions (lower variance) • Less personalized results (higher bias) • Works better for new users (not enough history) • More likely to converge General issues: • KNN is time consuming, doesn't scale well • Higher effect of "rich-get-richer" for popular items
  • 16. Also referred as Latent factor methods Approaches: 1. SVD 2. Matrix factorization 3. Neural Networks Model-based CF
  • 17. Model-based CF: Matrix Factorization Image source Matrix Factorization
  • 18. Training: • Initialize P and Q matrices with small random numbers • Teach P and Q • Alternating Least Squares • Stochastic Gradient Descent Predictions: MF algorithm
  • 19. MF example Latent features are calculated via MF: User-item score is the dot product: Item-item similarity is the cosine similarity:
  • 20. • MF with Biases: handling bias of some users giving higher ratings than others • MF with Side Features: adding data to handle the cold start problem (i.e. user occupation) • MF with Temporal Features: handling temporal changes of the data (i.e. occupation change) • Factorization Machine: extra item features + higher order interactions • MF with Fixture of tastes: give users several tastes • Variational MF MF improvements Source article
  • 21. Why? • Collaborative filtering uses robust approach based on similarities between customer tastes and can make cross-genre(category) recommendations • Pure CF lacks information about items and suffer from sparsity in data • Content-based filtering may find deep similarities between items and suggest novel and surprising new items • Only CBF can bring value from images, sounds, texts • All modern implementations are hybrid Hybrid More info: https://www.researchgate.net/publication/263377228_Hybrid_Recommender_Systems_Survey_and_Experiments
  • 24. • Modeling the non-linear interactions in the data • Feature extraction directly from the content: • Image, audio, text • Ability to use heterogeneous data (interactions, content) in the same model • Powerful for sequential modeling tasks (next item prediction, session-based recommendations) • Better representation learning of users and items for CF Deep Learning for RecSys
  • 25. Bringing NNs to CF Neural Collaborative Filtering Deep Factorization Machine Several more NN-based improvements can be found here: https://towardsdatascience.com/recsys-series-part-5-neural- matrix-factorization-for-collaborative-filtering-a0aebfe15883
  • 26. • Prior to Deep Learning approach used Matrix Factorization • 2 Neural Networks: • Candidates generation: broad personalization via CF (retrieves only couple of hundreds of videos) • Ranking: assigns a score to each video using a rich set of features from item and user • Uses implicit signals (full watching is positive), not explicit (thumbs up/down) • Use video "age" as a feature to recommend more new content • Rely mostly on A/B testing results https://research.google/pubs/pub45530/ Youtube recommendations
  • 27. • Uses implicit customer feedback • Back in 2014 used Weighted MF • Used MFCC and CNNs to extract features from music – helps with cold start and finding unpopular tracks • Used NLP to extract features from song texts and other textual information • Basically uses a hybrid recommender system https://benanne.github.io/2014/08/05/spotify-cnns.html https://www.oreilly.com/radar/personalization-of-spotify-home- and-tensorflow/ Spotify recommendations
  • 28. • Autoencoder-based recommendations: • to learn the lower-dimensional feature representations at the bottleneck layer • to fill in the blanks of the user-item interaction matrix directly in the reconstruction layer • CNN-based recommendations: • to extract features from images • to extract features from audio and video • to extract features from texts • RNN-based recommendations • to extract sequential patterns in session-based tasks • and many others: https://towardsdatascience.com/recommendation-system-series-part-2-the-10- categories-of-deep-recommendation-systems-that-189d60287b58 Other DL approaches
  • 29. Metrics: • When predicting rating: MSE, RMSE, etc. • When predicting binary output: Accuracy, Precision, Recall, F1 • When predicting top N: MAP@N, MAR@N • Coverage: % of items which are recommended (average CF predicts ~8-10%) Real-world testing: • CTR or CR (product, ads, etc.) • Avg Time spent or Customer retention (content) • A/B testing • Serendipity: pleasant surprise, unintended discovery Evaluation
  • 30. • Novelty • Diversity • Serendipity • Interpretability • Adaptation Qualities of a good recommender
  • 31. • https://towardsdatascience.com/introduction-to-recommender-systems-6c66cf15ada • https://towardsdatascience.com/recommendation-system-series-part-1-an-executive-guide-to- building-recommendation-system-608f83e2630a • https://medium.com/libreai/a-glimpse-into-deep-learning-for-recommender-systems- d66ae0681775 • ADD: https://towardsdatascience.com/modern-recommender-systems-a0c727609aa8 Materials