Recommender Systems

Course outline and learning expectations

This is a tutor-led course. The outline shows what your tutor will cover; teaching materials and examinations are provided directly to enrolled students.

What you will learn

• Explain how recommender systems work and identify suitable African digital use cases
• Prepare explicit and implicit user-item interaction data for recommendation tasks
• Build user-based and item-based collaborative filtering models
• Build content-based recommenders using item features, TF-IDF and cosine similarity
• Combine collaborative and content signals in a hybrid recommender
• Apply practical strategies for user, item and system cold-start problems
• Evaluate recommendations using relevance, ranking, coverage and diversity metrics
• Build, test, document and present a working Python recommender system responsibly

Modules

1. 1

   Foundations and African applications

   Understand recommendation problems and identify responsible applications in Uganda, East Africa and Africa

   Users, items and interactionsExplicit and implicit feedbackCommon recommendation tasksApplications in e-commerce, education, media and tourismResponsible personalisation
2. 2

   Collaborative filtering basics

   Build neighbourhood-based recommendations from user-item interaction data

   User-item matricesUser-based collaborative filteringItem-based collaborative filteringSimilarity measuresTop-N recommendationsSparsity and popularity bias
3. 3

   Content-based filtering

   Recommend items using descriptions, categories and other item attributes

   Feature engineeringCategorical featuresTF-IDF for textCosine similarityUser profilesExplaining recommendations
4. 4

   Hybrid recommendation approaches

   Combine collaborative, content and rule-based signals to improve practical recommendations

   Weighted hybridsSwitching methodsCascade methodsCandidate generation and rankingBusiness rulesComparing hybrid designs
5. 5

   Cold-start and limited data

   Develop useful recommendation strategies when users, items or interactions are new or limited

   User cold startItem cold startPopularity and segment baselinesPreference onboardingCollecting data with consentSparse-data African contexts
6. 6

   Evaluating recommendation quality

   Measure whether recommendations are relevant, useful and sufficiently varied

   Train, validation and test splitsPrecision at KRecall at KMAP and NDCGRMSE for rating predictionCoverage, diversity and serendipityOffline and online evaluation concepts
7. 7

   Building a recommender with Python

   Create an end-to-end prototype from data preparation to recommendation output

   Data cleaning with pandasExploratory analysisCollaborative model implementationContent model implementationGenerating and explaining recommendationsTesting and documentation
8. 8

   Capstone — What to Study Next Recommender

   Build and present a working student recommender based on performance and topic history

   Problem definitionDataset designModel selectionRecommendation logicEvaluationPrivacy and fairness reviewDocumentation and demonstration