Learn to design, build, and evaluate recommender systems for commerce and content. Advanced training in developing recommender systems to improve the user experience.
Suggested by: Coursera (What is Coursera?)
No prior knowledge required
No unnecessary risks
This course, intended to serve as a first course in recommender system skills, introduces the concept of recommender systems, reviews various examples in detail, and guides you through unoptimized recommendation using summary statistics and product associations, stereotype or demographic-based recommendations, and content-based recommendations. After completing the course, you will be able to calculate a variety of recommendations from data using basic spreadsheet tools, and if you complete the special track, you will also program these recommendations using the open-source recommendation tool LensKit.
In addition to detailed lectures and interactive exercises, the course includes interviews with several leaders in research and practice on advanced topics and current directions in recommender systems.
In this course, you will learn the basic techniques for personalized recommendations using nearest neighbor techniques. First, you will learn about cross-user collaborative filtering, an algorithm that identifies other people with similar tastes to a target user and combines their ratings to recommend products to that user. You will explore and implement variations of this algorithm, and discover the advantages and disadvantages of the general approach. Next, you will learn about cross-item collaborative filtering, an algorithm that identifies general associations between products from user ratings, but uses these associations to provide personalized recommendations based on the user’s own ratings.
In this course, you will learn how to evaluate recommender systems. You will be introduced to several families of metrics, including metrics for measuring predictive accuracy, ranking accuracy, decision support, and other metrics such as variety, product coverage, and surprises. You will understand how different metrics relate to different user goals and business objectives. You will also learn how to perform offline evaluations (i.e., how to prepare and sample data, and how to aggregate results). You will also learn about online (experimental) evaluation. Upon completion of the course, you will have the tools to compare different recommender system options for a wide range of uses.
In this course, you will learn a variety of hybrid matrix decomposition and machine learning techniques for recommender systems. Starting with basic matrix decomposition, you will understand both the conceptual and practical details of building recommender systems based on reducing the dimensionality of the user and product preference space. Then, you will learn about techniques that combine the advantages of different algorithms into powerful hybrid recommenders.
This capstone project course for recommender system skills brings together everything you’ve learned about recommender system algorithms and evaluation in a comprehensive project on recommender analysis and design. You’ll be given a case study in which you’ll need to select and justify the design of a recommender system by analyzing the recommendation goals and algorithm performance. Learners in the special track will focus on experimentally evaluating the algorithms against medium-sized datasets. The standard track will have a mix of vendor results and spreadsheet exploration.
Both groups created a final report documenting the analysis, the chosen solution, and the reasoning for that solution.
