Training and testing of recommender systems on data missing not at random

@inproceedings{Steck2010TrainingAT,
  title={Training and testing of recommender systems on data missing not at random},
  author={Harald Steck},
  booktitle={KDD},
  year={2010}
}
Users typically rate only a small fraction of all available items. We show that the absence of ratings carries useful information for improving the top-k hit rate concerning all items, a natural accuracy measure for recommendations. As to test recommender systems, we present two performance measures that can be estimated, under mild assumptions, without bias from data even when ratings are missing not at random (MNAR). As to achieve optimal test results, we present appropriate surrogate… CONTINUE READING

Similar Papers

Figures, Tables, Results, and Topics from this paper.

Key Quantitative Results

  • Depending on the value of k, our approach resulted in a 39 - 50% higher top-k hit rate compared to state-of-the-art recommender systems in our experiments on the Net.ix data.

Explore Further: Topics Discussed in This Paper

Citations

Publications citing this paper.
SHOWING 1-10 OF 132 CITATIONS, ESTIMATED 92% COVERAGE

l-Injection: Toward Effective Collaborative Filtering Using Uninteresting Items

  • IEEE Trans. Knowl. Data Eng.
  • 2019
VIEW 9 EXCERPTS
CITES BACKGROUND & RESULTS
HIGHLY INFLUENCED

Improving the accuracy of top-N recommendation using a preference model

  • Inf. Sci.
  • 2016
VIEW 11 EXCERPTS
CITES BACKGROUND & METHODS
HIGHLY INFLUENCED

Self-training Temporal Dynamic Collaborative Filtering

VIEW 19 EXCERPTS
CITES BACKGROUND & METHODS
HIGHLY INFLUENCED

Evaluation of recommendations: rating-prediction and ranking

  • RecSys
  • 2013
VIEW 19 EXCERPTS
CITES METHODS & BACKGROUND
HIGHLY INFLUENCED

Monte Carlo Estimates of Evaluation Metric Error and Bias

VIEW 4 EXCERPTS
CITES BACKGROUND & RESULTS
HIGHLY INFLUENCED

Collaborative Filtering-Based Recommendation of Online Social Voting

  • IEEE Transactions on Computational Social Systems
  • 2017
VIEW 11 EXCERPTS
CITES BACKGROUND & METHODS
HIGHLY INFLUENCED

Recommender Systems

  • Springer International Publishing
  • 2016
VIEW 5 EXCERPTS
CITES BACKGROUND & METHODS
HIGHLY INFLUENCED

“Told you i didn't like it”: Exploiting uninteresting items for effective collaborative filtering

  • 2016 IEEE 32nd International Conference on Data Engineering (ICDE)
  • 2016
VIEW 5 EXCERPTS
CITES RESULTS, BACKGROUND & METHODS
HIGHLY INFLUENCED

FILTER CITATIONS BY YEAR

2010
2019

CITATION STATISTICS

  • 33 Highly Influenced Citations

  • Averaged 16 Citations per year over the last 3 years

References

Publications referenced by this paper.