def sample_exp():
# step 1: binary classification
libfm = LibFM()
db = Database('movielens', 718, 8928, 'ratings.csv', 'movies.attr')
db.load_data()
db.make_train_test_matrix()
db.add_negative_data()
db.dump_libfm_data('train_step1.libfm', 'test_step1.libfm')
def sample_exp():
# making data from origin format to libfm format
#db = Database('movielens', 718, 8928, 'ratings.csv', 'movies.attr')
db = Database('amazon', 1000, 5000, 'cands_book.rating', 'cands_book.item')
db.load_data()
db.make_train_test_matrix()
libfm = LibFM()
# step 1: classification to predict interest
db.dump_libfm_data('train_step1.libfm', 'test_step1.libfm', add_negative=True, binary=True)
libfm.run('c', 'train_step1.libfm', 'test_step1.libfm', 'pred.libfm', iter_num=1)
step1_pred_list = db.load_pred_list('pred.libfm', 'c')
# TODO: merge accuracy accessing into libfm.py and call it here
# step 2: regression to predict possible rating
db.dump_libfm_data('train_step2.libfm', 'test_step2.libfm', add_negative=False, binary=False)
libfm.run('r', 'train_step1.libfm', 'test_step1.libfm', 'pred.libfm', iter_num=1)
step2_pred_list = db.load_pred_list('pred.libfm', 'r')
