# Matrix Factorization For Recommendation

How to recommend items based on rating data?

## Data

Movielens-100k (Movie review)


## Matrix Factorization



## What is good baseline?


1. Uniform Random (0~5)
2. Global Means
= (2+4+5+1+2+1+4)/7
= 3.28
3. Mean of means
= (Mean + item_mean + person_mean)/3
= (3.28+ 5 + (1+4)/2)/3
= 3.6

## Used library

Tried method:
PMF, PMF MCMC, SVD, SVD ++

1. PMF MAP -> Bayesian approach
2. PMF MCMC -> Bayesian + Markov chain monte carlo　approach
3. Funk SVD -> Netflix prize algorithm 3rd place
4. SVD ++ -> Funk SVD + implicit rating

### Error
RMSE = (1/N * sum((Rating - predicted rating)**2)) **0.5

# Result

### Tutorial

### Data

### predict

### result

### Tried

ml-100k latest

### Data

### predict



###

### Data

### predict