Data Mining

In this assignment you are going to implement the SVD algorithm, that we have discussed in the class, for solving the rating prediction problem. The algorithm is also explained in section 5.3.1 of Recommender Systems Handbook. Apply this algorithm on both movielens 100k and 1m datasets. Use the same experiment techniques (e.g., cross validation) that you have used in assignment I. You should find the MAE results for both datasets. Also, for each dataset, plot a figure showing the relationship between the number of factors and MAE. You can try factor sizes such as 10, 20, ..., 100. To detect whether your algorithm has converged plot a graph which shows the cost function value as a function of the number of iterations. Group work is not allowed. Deadline: April 30, 2015. Link: [login to view URL]~tevfik/courses/cmp5101/[login to view URL]