Skip to content
Stanford CS 399: Improving Estimates of Item-Item Similarity & Making Inter-Retailer Recommendations
Branch: master
Clone or download
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Permalink
Type Name Latest commit message Commit time
Failed to load latest commit information.
amazon
bdales
cgi-bin
common
gap
macys
matching
paper
paper2
proposal
website
.gitignore
README.md
report.pdf
report2.pdf

README.md

Stanford CS 399 Independent Study Projects

Project 1: Improving Item-Item Similarity Estimation during Collaborative Filtering

Project 2: A Technique for Inter-Retailer Recommendations

Summary

Project 1

This project explores the problem of inaccurate item-item similarity estimates due to a lack of data within the context of collaborative filtering. Two novel item-item similarity algorithms are implemented that leverage large sets of user-item rating data. The first is based on a model of noisy similarity score generation and is effective at estimating both future Pearson correlations and cosine similarities. The second models noisy ratings as generated by similarity score dependent distributions. Using the notion of user-predictvity, accurate estimates of future Pearson correlation are made. See report.pdf for details.

Project 2

This project explores the problem of making recommendations across retailers using only publicly available information. Several inter-retailer recommender algorithms are implemented along with an off-line experiment that compares the performance of each algorithm using real data. A novel latent feature recommender is implemented that leverages public intra-retailer recommendation information. See report2.pdf for details.

You can’t perform that action at this time.
You signed in with another tab or window. Reload to refresh your session. You signed out in another tab or window. Reload to refresh your session.