Chapter 9

Title: Constructing Recommender Systems in RapidMiner


Chapter 9 introduces the RapidMiner Extension for Recommender Systems. This extension allows building more sophisticated recommendation systems than described in the previous chapter. The application task in this chapter is to recommend appropriate video lectures to potential viewers. The recommendations can be based on the content of the lectures or on the viewing behavior or on both. The corresponding approaches are called content-based, collaborative, and hybrid recommendation, respectively. Content-based recommendations can be based on attributes or similarity and collaborative recommendation systems deploy neighborhoods or factorization. This chapter explains, evaluates, and compares these approaches. It also demonstrates how to make RapidMiner processes available as RapidAnalytics web services, i.e., how to build a recommendation engine and make it available for real-time recommendations and easy integration into web sites, online shops, and other systems via web services.

Table of Contents

9.1 Introduction
9.2 The Recommender Extension
9.2.1 Recommendation Operators
9.2.2 Data Format
9.2.3 Performance Measures
9.3 The Dataset
9.4 Collaborative-based Systems
9.4.1 Neighbourhood-based Recommender Systems
9.4.2 Factorization-based Recommender Systems
9.4.3 Collaborative Recommender Workflows
9.4.4 Iterative Online Updates
9.5 Content-based Recommendation
9.5.1 Attribute-based Content Recommendation
9.5.2 Similarity-based Content Recommendation
9.6 Hybrid Recommender Systems
9.7 Providing RapidMiner Recommender System Workflows as Web Services Using RapidAnalytics
9.7.1 Simple Recommender System Web Service
9.7.2 Guidelines for Optimizing Workflows for Service Usage
9.8 Summary
9.8 Glossary
9.8 Bibliography

Dataset & Processes: Click here to download