Part 1: SVD&SVD++
- Implemented SVD and SVD++ using Multiplicative update rules
- Used KL divergence and Euclidean distance to compute cost, respectively.
Part 2: Movie Recommender
-
Developed a recommender system based on the SVD++ algorithm to predict users’ preferences for unseen movies based on their similarity to other users
-
The dataset contains 100,000 ratings from 1000 users on 1700 movies. Data Source: https://grouplens.org/datasets/movielens/100k/