MyMediaLite rating prediction 2.02
usage: rating_prediction --training-file=FILE --recommender=METHOD [OPTIONS]
recommenders (plus options and their defaults):
- SocialMF num_factors=10 regularization=0.015 social_regularization=1 learn_rate=0.01 num_iter=30 init_mean=0 init_stddev=0.1 (needs --user-relations=FILE)
- MultiCoreMatrixFactorization num_factors=10 bias_reg=0.0001 reg_u=0.015 reg_i=0.015 learn_rate=0.01 num_iter=30 bold_driver=True init_mean=0 init_stddev=0.1 optimize_mae=False num_blocks=100
- BiPolarSlopeOne
- BiasedMatrixFactorization num_factors=10 bias_reg=0.0001 reg_u=0.015 reg_i=0.015 learn_rate=0.01 num_iter=30 bold_driver=False init_mean=0 init_stddev=0.1 optimize_mae=False
- FactorWiseMatrixFactorization num_factors=10 shrinkage=25 sensibility=1E-05 init_mean=0 init_stdev=0.1 num_iter=10
- GlobalAverage
- ItemAttributeKNN k=inf reg_u=10 reg_i=5 (needs --item-attributes=FILE)
- ItemAverage
- ItemKNNCosine k=inf reg_u=10 reg_i=5
- ItemKNNPearson k=inf shrinkage=10 reg_u=10 reg_i=5
- MatrixFactorization num_factors=10 regularization=0.015 learn_rate=0.01 num_iter=30 init_mean=0 init_stddev=0.1
- SlopeOne
- UserAttributeKNN k=inf reg_u=10 reg_i=5 (needs --user-attributes=FILE)
- UserAverage
- UserItemBaseline reg_u=15 reg_i=10 num_iter=10
- UserKNNCosine k=inf reg_u=10 reg_i=5
- UserKNNPearson k=inf shrinkage=10 reg_u=10 reg_i=5
- TimeAwareBaseline num_iter=30 bin_size=70 beta=0.4 user_bias_learn_rate=0.003 item_bias_learn_rate=0.002 alpha_learn_rate=1E-05 item_bias_by_time_bin_learn_rate=5E-06 user_bias_by_day_learn_rate=0.0025 user_scaling_learn_rate=0.008 user_scaling_by_day_learn_rate=0.002 reg_u=0.03 reg_i=0.03 reg_alpha=50 reg_item_bias_by_time_bin=0.1 reg_user_bias_by_day=0.005 reg_user_scaling=0.01 reg_user_scaling_by_day=0.005
- TimeAwareBaselineWithFrequencies num_iter=40 bin_size=70 beta=0.4 user_bias_learn_rate=0.00267 item_bias_learn_rate=0.000488 alpha_learn_rate=3.11E-06 item_bias_by_time_bin_learn_rate=1.15E-06 user_bias_by_day_learn_rate=0.000257 user_scaling_learn_rate=0.00564 user_scaling_by_day_learn_rate=0.00103 reg_u=0.0255 reg_i=0.0255 reg_alpha=3.95 reg_item_bias_by_time_bin=0.0929 reg_user_bias_by_day=0.00231 reg_user_scaling=0.0476 reg_user_scaling_by_day=0.019 frequency_log_base=6.76 item_bias_at_frequency_learn_rate=0.00236 reg_item_bias_at_frequency=1.1E-08
- CoClustering num_user_clusters=3 num_item_clusters=3 num_iter=30
method ARGUMENTS have the form name=value
general OPTIONS:
--recommender=METHOD set recommender method (default: BiasedMatrixFactorization)
--recommender-options=OPTIONS use OPTIONS as recommender options
--help display this usage information and exit
--version display version information and exit
--random-seed=N initialize the random number generator with N
--rating-type=float|byte|double store ratings internally as floats or bytes or doubles (default)
files:
--training-file=FILE read training data from FILE
--test-file=FILE read test data from FILE
--file-format=movielens_1m|kddcup2011|ignore_first_line|default
--data-dir=DIR load all files from DIR
--user-attributes=FILE file containing user attribute information, 1 tuple per line
--item-attributes=FILE file containing item attribute information, 1 tuple per line
--user-relations=FILE file containing user relation information, 1 tuple per line
--item-relations=FILE file containing item relation information, 1 tuple per line
--save-model=FILE save computed model to FILE
--load-model=FILE load model from FILE
prediction options:
--prediction-file=FILE write the rating predictions to FILE
--prediction-line=FORMAT format of the prediction line; {0}, {1}, {2} refer to user ID, item ID,
and predicted rating, respectively; default is {0}\\t{1}\\t{2}
evaluation options:
--cross-validation=K perform k-fold cross-validation on the training data
--show-fold-results show results for individual folds in cross-validation
--test-ratio=NUM use a ratio of NUM of the training data for evaluation (simple split)
--chronological-split=NUM|DATETIME use the last ratio of NUM of the training data ratings for evaluation,
or use the ratings from DATETIME on for evaluation (requires time information
in the training data)
--online-evaluation perform online evaluation (use every tested rating for incremental training)
--search-hp search for good hyperparameter values (experimental)
--compute-fit display fit on training data
options for finding the right number of iterations (iterative methods)
--find-iter=N give out statistics every N iterations
--max-iter=N perform at most N iterations
--epsilon=NUM abort iterations if RMSE is more than best result plus NUM
--rmse-cutoff=NUM abort if RMSE is above NUM
--mae-cutoff=NUM abort if MAE is above NUM
[% INCLUDE footer %]