Community-Based Link - Complex Networks - Lecture Slides, Slides of Computer Networks

Download Community-Based Link - Complex Networks - Lecture Slides and more Slides Computer Networks in PDF only on Docsity!

Community-Based Link Prediction/Recommendation in the Bipartite Network of BoardGameGeek.com

Introduction

The Problem

Related Works

Conclusion

Questions / Comments

• Very important for online businesses
• Drive demand for product
• Companies have had contests with

million dollar prizes to increase performance

Recommender Systems

Introduction

The Problem

Related Works

Conclusion

Questions / Comments

• Users & Item Profiles
• Based on content (e.g. genre, demographics, length, etc.)

Content Based

• Users & Items similar to those in the past
• More abstract, only links matter

Collaborative

Based

User

User

Item

Item

• Memory-based
  • Use entire dataset directly
• Model-based
  • Create a model based on data
  • Uses model to make recommendations

Collaborative Filtering

J. S. Breese, et al., "Empirical analysis of predictive algorithms for collaborative filtering," 1998

• Recommendations are based on the

users that have liked items similar to

ones the user has liked in the past

User-based Collaborative Filtering

• Recommendations are based on the items

rated/bought similarly to other items

Item-based Collaborative Filtering

• Sparsity is an issue
• Consumer-product matrix looks like:
• Instead, represent the matrix as a bipartite graph
• Significantly better results under sparse conditions
• Computationally expensive

Link-analysis approach

Z. Huang , et al. , "A Link analysis approach to recommendation under sparse data," 2004.

Link-analysis approach

Z. Huang , et al. , "A Link analysis approach to recommendation under sparse data," 2004.

• CF Performs poorly for “cold-start” users
• Trust-based recommenders work well if a user is at least connected to a large component
• Sparsity forces a trust-based approach to consider weakly trusted neighbors
• Added a random walk model to allow for defining and measuring a confidence metric
• Protects agains things like faked profiles or spammed ratings

TrustWalker

M. Jamali and M. Ester, "TrustWalker: a random walk model for combining trust-based and item-based recommendation," 2009

• Require large amount of knowledge about users and items
• Often use textural information (website recommenders)
• Explicit or implicit profile generation
• Can over specialize (some workarounds)

Content-Based Filtering

G. Adomavicius and A. Tuzhilin, "Toward the next generation of recommender systems: A survey of the state-of-the-art and possible extensions,"

• Weighted
• Switched
• Mixed
• Feature combination
• Cascade

Methods of Hybrid Filtering

R. Burke, "Hybrid recommender systems: Survey and experiments,"

Clustering Approach

Q. Li and B. M. Kim, "Clustering approach for hybrid recommender system," 2003