Skip to main content

Recommender System Utilities

Project description

Recommender Utilities

This package (reco_utils) contains functions to simplify common tasks used when developing and evaluating recommender systems. A short description of the sub-modules is provided below. For more details about what functions are available and how to use them, please review the doc-strings provided with the code.

See the online documentation.


The AzureML submodule contains utilities to train, tune and operationalize recommendation systems at scale using AzureML.


This submodule contains high-level utilities for defining constants used in most algorithms as well as helper functions for managing aspects of different frameworks: gpu, spark, jupyter notebook.


Dataset includes helper functions for interacting with Azure Cosmos databases, pulling different datasets and formatting them appropriately as well as utilities for splitting data for training / testing.

Data Loading

There are dataloaders for several datasets. For example, the movielens module will allow you to load a dataframe in pandas or spark formats from the MovieLens dataset, with sizes of 100k, 1M, 10M, or 20M to test algorithms and evaluate performance benchmarks.

df = movielens.load_pandas_df(size="100k")

Splitting Techniques

Currently three methods are available for splitting datasets. All of them support splitting by user or item and filtering out minimal samples (for instance users that have not rated enough item, or items that have not been rated by enough users).

  • Random: this is the basic approach where entries are randomly assigned to each group based on the ratio desired
  • Chronological: this uses provided timestamps to order the data and selects a cut-off time that will split the desired ratio of data to train before that time and test after that time
  • Stratified: this is similar to random sampling, but the splits are stratified, for example if the datasets are split by user, the splitting approach will attempt to maintain the same set of items used in both training and test splits. The converse is true if splitting by item.


The evaluation submodule includes functionality for performing hyperparameter sweeps as well as calculating common recommender metrics directly in python or in a Spark environment using pyspark.

Currently available metrics include:

  • Root Mean Squared Error
  • Mean Absolute Error
  • R2
  • Explained Variance
  • Precision at K
  • Recall at K
  • Normalized Discounted Cumulative Gain at K
  • Mean Average Precision at K
  • Area Under Curve
  • Logistic Loss


The recommender submodule contains implementations of various algorithms that can be used in addition to external packages to evaluate and develop new recommender system approaches. A description of all the algorithms can be found on this table. Next a list of the algorithm utilities:

  • Cornac
  • DeepRec (includes xDeepFM and DKN)
  • FastAI
  • LightGBM
  • NCF
  • NewsRec (includes LSTUR, NAML NPA and NRMS)
  • RBM
  • SAR
  • Surprise
  • Vowpal Wabbit (VW)
  • Wide&Deep


This submodule contains utilities for performing hyperparameter tuning.

Project details

Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Source Distribution

recommender-utils-2021.2.post1623854186.tar.gz (144.5 kB view hashes)

Uploaded source

Built Distribution

Supported by

AWS AWS Cloud computing and Security Sponsor Datadog Datadog Monitoring Fastly Fastly CDN Google Google Download Analytics Microsoft Microsoft PSF Sponsor Pingdom Pingdom Monitoring Sentry Sentry Error logging StatusPage StatusPage Status page