In this notebook we will implement recommendation system based on Slope-One and SVD algorithms
The original paper of Slope-one algorithm
We will explore Movielens 100k dataset and compare the accuracy rate of the recommendation system built with 2 techiques:
- The only Slop-One
- SVD (low rank approximation) + Slope-One on the reduced matrix (The idea is taken from this paper)
In order to see graphs you need to launch notebook localy or use links from output_data/graphs. This is because GitHub can't show html in jupyter notebook