Skip to content

fuhailin/Memory-based-collaborative-filtering

Repository files navigation

Memory-based-collaborative-filtering

Build Status

Contain User-based CF(UBCF),Item-based CF(IBCF) A robust k-nearest neighbors Recommender System use MovieLens dataset in Python

User-based collaborative filter

K=25   RunTime:1s RMSE:0.940611 MAE:0.884748.

image image

Memory-based algorithms are easy to implement and produce reasonable prediction quality. The drawback of memory-based CF is that it doesn’t scale to real-world scenarios and doesn’t address the well-known cold-start problem, that is when new user or new item enters the system.

About

高效的协同过滤算法实现——推荐系统

Topics

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages