search

recommenders / rival

RiVal recommender system evaluation toolkit
rival.recommenders.net
Apache License 2.0
151 stars 40 forks source link

Support for hybrid data input #68

Closed alansaid closed 10 years ago

alansaid commented 10 years ago

The data model should support richer data sets (e.g. more than 4 columns tsv/csv files used as input).

Probably good idea to wait till https://commons.apache.org/proper/commons-csv/ reaches a stable release.

Related to #54

abellogin commented 10 years ago

Why is issue 54 related to this? I remember there was an issue about csv, but I cannot find it now.

alansaid commented 10 years ago

If we create a complete pipeline, the data model will need to be aware of the dataset it's based on, so it will have an effect on the complete process. (Granted we won't have to rewrite anything provided that we don't remove current methods in the datamodel)

abellogin commented 10 years ago

Ok, thanks. Thought it was a typo, but it makes sense.

alansaid commented 10 years ago

Apache Commons CSV 1.0 has now been released. I'll start working on adding this to a data reader. http://mail-archives.apache.org/mod_mbox/commons-user/201408.mbox/%3CCAB917R%2BYoXZr-8zftqZ4hgYqFz6cgRXxqqza4gtaF7LdrKMJQw%40mail.gmail.com%3E

alansaid commented 10 years ago

Apache Commons CSV 1.0 is now used in the parser in rival-core. Additional configurations (csv/tsv/json) can be added in rival-examples (see rival-examples/RandomMahoutIBRecommenderEvaluator)