A Mixture-of-Gaussians model for estimating the magic barrier of the recommender system
Graphical abstract
Section snippets
Code metadata
Permanent link to reproducible Capsule: https://doi.org/10.24433/CO.8965551.v1.
Related work
This section first reviews the definition of the rating system. Second, we present the problem statement for collaborative filtering (CF). Finally, we analyze existing approaches for estimating MGBR.
Table 1 lists the notations used in the paper.
The estimation model with MoG noise
In this section, we first describe the estimating framework. Second, we present a MoG-based noise assumption. Third, we estimate the noise parameters related to the computation of the MGBR. Finally, we estimate the MGBR based on conditional-probability computation.
Experiments
In this section, we present the experimental scheme to answer the questions of using the MoG model to calculate the magic barriers:
- (1)
Given a dataset, how many Gaussian distributions are needed to approximate the user uncertainty? and
- (2)
Are the magic barriers estimated by the new model close to the results of the state-of-the-art algorithms?
The answers to these problems help evaluate the rationality of the proposed model.
We repeat the calculation times using different initial settings of the
Conclusions and further works
We have proposed a MoG model to estimate the MGBR of recommender system in terms of MAE and RMSE. The parameter can be calculated directly. The model is validated by state-of-the-art recommender algorithms on the real-world datasets. Our estimated magic barriers can be used to determine how much room for improvement in existing algorithms and to evaluate the quality of the data collected by recommender system.
The estimation model takes discrete rating as input when calculating conditional
CRediT authorship contribution statement
Heng-Ru Zhang: Conceptualization, Methodology, Writing, Software. Jie Qian: Data preprocessing, Writing. Hui-Lin Qu: Validation, Editing. Fan Min: Conceptualization, Writing – reviewing.
Declaration of Competing Interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.
Acknowledgments
This work is supported in part by the National Natural Science Foundation of China (Grant 61902328), Natural Science Foundation of Sichuan Province, China (Grant 2019YJ0314), and Scientific Innovation Group for Youths of Sichuan Province, China (Grant 2019JDTD0017).
References (47)
- et al.
Sentiment based matrix factorization with reliability for recommendation
Expert Syst. Appl.
(2019) - et al.
A fuzzy model for managing natural noise in recommender systems
Appl. Soft Comput.
(2016) - et al.
Optimal classifiers with minimum expected error within a Bayesian framework-part I: Discrete and Gaussian models
Pattern Recognit.
(2013) - et al.
Regression-based three-way recommendation
Inform. Sci.
(2017) - et al.
Correcting noisy ratings in collaborative recommender systems
Knowl.-Based Syst.
(2015) - et al.
A noise correction-based approach to support a recommender system in a highly sparse rating environment
Decis. Support Syst.
(2019) On the period length of the generalized lagrange algorithm
J. Number Theory
(1987)- et al.
Recommender systems: introduction and challenges
- et al.
I like it. i like it not: Evaluating user ratings noise in recommender systems
- et al.
Users and noise: The magic barrier of recommender systems
Human error: models and management
Bmj
Neural systems responding to degrees of uncertainty in human decision-making
Science
Coherence and inconsistencies in rating behavior: estimating the magic barrier of recommender systems
User Model User-Adapt Int
Shilling attacks against recommender systems: a comprehensive survey
Artif. Intell. Rev.
Estimating the magic barrier of recommender systems: a user study
How to process uncertainty in machine learning?
Confident learning: Estimating uncertainty in dataset labels
J. Artificial Intelligence Res.
Aleatoric and epistemic uncertainty in machine learning: an introduction to concepts and methods
Mach. Learn.
Bayesian uncertainty estimation for batch normalized deep networks
Magic barrier estimation models for recommended systems under normal distribution
Appl. Intell.
Evaluating collaborative filtering recommender systems
ACM Trans. Inf. Syst.
Rating support interfaces to improve user experience and recommender accuracy
The magic barrier of recommender systems–no magic, just ratings
Cited by (4)
Lower bound estimation of recommendation error through user uncertainty modeling
2023, Pattern RecognitionCitation Excerpt :However, the inaccuracy of these artificially specified parameters generates additional uncertainty. Theoretical methods constitute SG [8] or MoG [9] models for obtaining the rating data and followed by a probabilistic model to estimate the MGBR. However, real-world uncertainty comprises a mixture of super- and sub-Gaussian components.
An effective explainable food recommendation using deep image clustering and community detection
2022, Intelligent Systems with ApplicationsGAME: GAussian Mixture Error-based meta-learning architecture
2023, Neural Computing and ApplicationsA trilevel analysis of uncertainty measuresin partition-based granular computing
2023, Artificial Intelligence Review
The code (and data) in this article has been certified as Reproducible by Code Ocean: (https://codeocean.com/). More information on the Reproducibility Badge Initiative is available at https://www.elsevier.com/physical-sciences-and-engineering/computer-science/journals.