Learners : Matchbox recommender : Command-line runners

Predictors

The Matchbox recommender in Infer.NET supports 4 modes of prediction - rating prediction, item recommendation, find related users, and find related items. There is a command-line runner for each one of these.

Rating predictor

Rating prediction is performed using the PredictRatings argument to Learner Recommender. It takes in a test set and a model, and produces a file of predictions.

Required parameters

Example

Learner Recommender PredictRatings --data TestSet.dat 
                                   --model TrainedRecommender.bin 
                                   --predictions RatingPredictions.dat

Item recommender

Item recommendation is performed using the RecommendItems argument to Learner Recommender. It takes in a test set and a model, and produces a file of predictions. Note that items are recommended with the intention to be later evaluated, so this module follows the prediction procedure explained here. That is, items are recommended for each unique user in the test set, and predictions are restricted to the items that each user has rated.

Required parameters

Optional parameters

  • max-items - maximum number of items to recommend; defaults to 5
  • min-pool-size - minimum size of the recommendation pool for a single user; defaults to 5

Example

Learner Recommender RecommendItems --data TestSet.dat 
                                   --model TrainedRecommender.bin
                                   --predictions ItemRecommendations.dat 
                                   --max-items 10

Related user finder

Related users can be found using the FindRelatedUsers argument to Learner Recommender. It takes in a test set and a model, and produces a file of predictions. Note that related users are found with the intention to be later evaluated, so this module follows the prediction procedure explained here. That is, related users are found for users who have rated a given number of items in common.

Required parameters

Optional parameters

  • max-users - maximum number of related users for a single user; defaults to 5
  • min-common-items - minimum number of items that the query user and the related user should have rated in common; defaults to 5
  • min-pool-size - minimum size of the related user pool for a single user; defaults to 5

Example

Learner Recommender FindRelatedUsers --data TestSet.dat
                                     --model TrainedRecommender.bin
                                     --predictions RelatedUsers.dat
                                     --max-users 10
                                     --min-common-items 2

Related item finder

Related items can be found using the FindRelatedItems argument to Learner Recommender. It takes in a test set and a model, and produces a file of predictions. Note that related items are found with the intention to be later evaluated, so this module follows the prediction procedure explained here. That is, related items are found for items which have been rated by a given number of users in common.

Required parameters

Optional parameters

  • max-items - maximum number of related items for a single item; defaults to 5
  • min-common-users - minimum number of users that the query item and the related item should have been rated by in common; defaults to 5
  • min-pool-size - minimum size of the related item pool for a single item; defaults to 5

Example

Learner Recommender FindRelatedItems --data TestSet.dat 
                                     --model TrainedRecommender.bin
                                     --predictions RelatedItems.dat
                                     --max-items 10 
                                     --min-common-users 2

©2009-2015 Microsoft Corporation. All rights reserved.  Terms of Use | Trademarks | Privacy Statement