We are a family owned USA based Corporate Training Company determined to help professionals, teams, and organizations improve.

Let us help you find the training program you are looking for.

If you can't find what you are looking for, contact us, we'll help you find it. We have over 800 training programs to choose from.

banner-img
Home » Technology » Big Data & Data Science » Practical Recommender Systems

Course Skill Level:

Foundational

Course Duration:

4 day/s

  • Course Delivery Format:

    Live, instructor-led.

  • Course Category:

    Big Data & Data Science

  • Course Code:

    PRECOSL21E09

Who should attend & recommended skills:

Python experienced developers with basic IT, higher math and statistics, Linux skills

Who should attend & recommended skills

  • Python experienced developers with infrastructure needed to get them up and running and recommendation systems, analysts, or others who are intending be able to read code in a programming language such as Python or Java. You should understand an SQL query, and have a basic understanding of higher math and statistics. Managers will find this course useful to get an overview of what a recommender system is and how it can be used in practice.
  • Skill-level: Foundation- Practical Recommender Systems for Intermediate skilled team members. This is not a basic class.
  • IT skills: Basic to Intermediate (1-5 years’ experience)
  • Higher math and statistics: Basic (1-2 years’ experience)
  • Linux: Basic, including familiarity with command-line options such as ls, cd, cp, and su
  • The ability to read code in a programming language such as Python or Java
  • An understanding of an SQL query
  • Attendees without a programming background like Python may view labs as follow along exercises or team with others to complete them.

About this course

Are you envious when Amazon recommends its products or when Netflix is spot-on with a recommendation for a user? Then here’s your chance to learn how to add these skills to your repertoire. Reading this course will give you an understanding of what recommender systems are and how to apply them in practice. To make a recommender work, many things need to perform in concert. You need to understand how to collect data from your users and how to interpret it, and you need a toolbox of different recommender algorithms so you can choose the best one for your particular scenario. Most importantly, you need to understand how to evaluate whether your recommender system is doing its job well. All this and more is hidden within this course.

Skills acquired & topics covered

  • Figures and code listings that explain concepts can get you only so far.
  • How to understand users and their behavior, and covers ways to collect data from users
  • Introduction to web analytics and how you can implement a dashboard where you can keep track of your recommenders
  • How behavioral data can be transformed into ratings
  • The problem of new users and products and simple solutions
  • Formulas for calculating similarity between users or content items such as movies
  • A way to mix types of recommenders
  • Introduction to ranking algorithms and methods for learning to rank recommendations.
  • Non-personalized recommendations

Course breakdown / modules

  • Real-life recommendations
  • Taxonomy of recommender systems
  • Machine learning and the Netflix Prize
  • The MovieGEEKs website
  • Building a recommender system

  • How (I think) Netflix gathers evidence while you browse
  • Finding useful user behavior
  • Identifying users
  • Getting visitor data from other sources
  • The collector
  • What users in the system are and how to model them

  • Why adding a dashboard is a good idea
  • Doing the analytics
  • Personas
  • MovieGEEKs dashboard

  • User-item preferences
  • Explicit or implicit ratings
  • Revisiting explicit ratings
  • What are implicit ratings?
  • Calculating implicit ratings
  • How to implement implicit ratings
  • Less frequent items provide more value

  • What’s a non-personalized recommendation?
  • How to make recommendations when you have no data
  • Implementing the chart and the groundwork for the recommender system component
  • Seeded recommendations

  • What’s a cold start?
  • Keeping track of visitors
  • Addressing cold-start problems with algorithms
  • Those who doesn’t ask, won’t know
  • Using association rules to start recommending things fast

  • Why similarity?
  • Essential similarity functions
  • k-means clustering
  • Implementing similarities

  • Collaborative filtering: A history lesson
  • Calculating recommendations
  • Calculating similarities
  • Amazon’s algorithm to precalculate item similarity
  • Ways to select the neighborhood
  • Finding the right neighborhood
  • Ways to calculate predicted ratings
  • Prediction with item-based filtering
  • Cold-start problems
  • A few words on machine learning terms
  • Collaborative filtering on the MovieGEEKs site
  • What’s the difference between association rule recs and collaborative recs?
  • Levers to fiddle with for collaborative filtering
  • Pros and cons of collaborative filtering

  • Business wants lift, cross-sales, up-sales, and conversions
  • Why is it important to evaluate?
  • How to interpret user behavior
  • What to measure
  • Before implementing the recommender…
  • Types of evaluation
  • Offline evaluation
  • Offline experiments
  • Implementing the experiment in MovieGEEKs
  • Evaluating the test set
  • Online evaluation
  • Continuous testing with exploit/explore

  • Descriptive example
  • Content-based filtering
  • Content analyzer
  • Extracting metadata from descriptions
  • Finding important words with TF-IDF
  • Topic modeling using the LDA
  • Finding similar content
  • Creating the user profile
  • Content-based recommendations in MovieGEEKs
  • Evaluation of the content-based recommender
  • Pros and cons of content-based filtering

  • Sometimes it’s good to reduce the amount of data
  • Example of what you want to solve
  • A whiff of linear algebra
  • Matrix
  • What’s factorization?
  • Constructing the factorization using SVD
  • Adding a new user by folding in
  • How to do recommendations with SVD
  • Baseline predictors
  • Temporal dynamic
  • Constructing the factorization using Funk SVD
  • Root Mean Squared Error
  • Gradient descent
  • Stochastic gradient descent
  • And finally, to the factorization
  • Adding biases
  • How to start and when to stop
  • Doing recommendations with Funk SVD
  • User vector
  • The items the user likes
  • Funk SVD implementation in MovieGEEKs
  • What to do with outliers
  • Keeping the model up to date
  • Faster implementation
  • Explicit vs. implicit data
  • Evaluation
  • Levers to fiddle with for Funk SVD

  • The confused world of hybrids
  • The monolithic
  • Mixing content-based features with behavioral data to improve collaborative filtering recommenders
  • Mixed hybrid recommender
  • The ensemble
  • Switched ensemble recommender
  • Weighted ensemble recommender
  • Linear regression
  • Feature-weighted linear stacking (FWLS)
  • Meta features: Weights as functions
  • The algorithm
  • Implementation

  • Learning to rank an example at Foursquare
  • Re-ranking
  • What’s learning to rank again?
  • The three types of LTR algorithms
  • Bayesian Personalized Ranking
  • Ranking with BPR
  • Math magic (advanced wizardry)
  • The BPR algorithm
  • BPR with matrix factorization
  • Implementation of BPR
  • Doing the recommendations
  • Evaluation
  • Levers to fiddle with for BPR

  • Algorithms
  • Context
  • Human-computer interactions
  • Choosing a good architecture
  • What’s the future of recommender systems?
  • User profiles
  • context
  • Algorithms
  • Privacy
  • Architecture
  • Surprising recommendations

Browse our programs to take the next step toward advancing yourself, your team, and organization.

Browse Courses