Build SciPy sparse matrices

sparse_matrices.py

def sparse_matrices(df):
    '''creates the sparse user-item and item-user matrices'''

    # using a scalar value (40) to convert ratings from a scale (1-5) to a like/click/view (1)
    alpha = 40

    sparse_user_item = csr_matrix( ([alpha]*len(df['movie_id']), (df['user_id'], df['movie_id']) ))
    # transposing the item-user matrix to create a user-item matrix
    sparse_item_user = sparse_user_item.T.tocsr()
    # save the matrices for recalculating user on the fly 
    save_npz("sparse_user_item.npz", sparse_user_item)
    save_npz("sparse_item_user.npz", sparse_item_user)

    return sparse_user_item, sparse_item_user