The Development of Topic Model Based on Beta-Negative Binomial Process

Ya Shu Liu; Han Bing Yan

doi:10.4028/www.scientific.net/AMM.427-429.1597

Paper Titles

Dynamic Spectrum Allocation Using Q-Learning in Cognitive Radio Systems
p.1579

Algorithm for Maneuvering Target Tracking Based on MATLAB
p.1585

An Image Fusion Algorithm Based on NSCT
p.1589

A Novel Image Fusion Algorithm
p.1593

The Development of Topic Model Based on Beta-Negative Binomial Process
p.1597

Mirror Transform to Conquer Deception for Particle Swarm Optimization
p.1601

Multi-Instance Learning for Image Retrieval with Relevance Feedback
p.1606

Affine Image Registration Based on Geometric Inference
p.1610

Bayesian Network Structure Learning Based on Probability Density Kernel Estimation
p.1614

HomeApplied Mechanics and MaterialsApplied Mechanics and Materials Vols. 427-429The Development of Topic Model Based on...

The Development of Topic Model Based on Beta-Negative Binomial Process

Abstract:

. Topic Model is one of the important subfields in Data Mining, which has been developed very quickly and has been applicated in many fields in recent years. Many researchers have been engaged in this field. In this paper, we introduce the BNB process based on Beta and Negative Binomial distribution, using the hierarchical distribution instead of Dirichlet in LDA. And we give the expression of parameter estimation used by Gibbs sampling. Then, BNB process is applicated in the text topic classification. We design experiments to decide the numbers of topics and compare the BNB process with LDA. Experiment results show that the BNB process has better performance over LDA in English Dataset, but they have almost the same result in Chinese micro-blog topic classification. Finally we analyze the problem and give the idea in further research.

You might also be interested in these eBooks

View Preview

Info:

Periodical:

Applied Mechanics and Materials (Volumes 427-429)

Pages:

1597-1600

DOI:

https://doi.org/10.4028/www.scientific.net/AMM.427-429.1597

Citation:

Cite this paper

Online since:

September 2013

Authors:

Ya Shu Liu*, Han Bing Yan

Keywords:

Beta Distribution, BNB Process, Latent Dirichlet Allocation (LDA), Negative Binomial Distribution, Topic Model

Export:

RIS, BibTeX

Price:

Permissions CCC:

Request Permissions

Permissions PLS:

Request Permissions

Сopyright:

Citation:

* - Corresponding Author

References

[1] J. Paisley, C. Wang and D.M. Blei. The Discrete Infinite Logistic Normal Distribution[J]. Bayesian Analysis. 2012, 7(4): 997-1034.

DOI: 10.1214/12-ba734

Google Scholar

[2] Xu Chen, Mingyuan Zhou and Lawrence Carin. The Contextual Focused Topic Model[C]. /Proceedings of the 18th ACM SIGKDD International Conference on Knowledge Discovery and DataMing. NY, USA: ACM Press, (2012).

DOI: 10.1145/2339530.2339549

Google Scholar

[3] O. E. Barndorff-Nielsen, D. G. Pollard, and N. Shephard. Integer-valued L´evy Processes and Low Latency Financial Econometrics[J]. Quantitative Finance. 2012, 12(4): 587-605.

DOI: 10.1080/14697688.2012.664935

Google Scholar

[4] L. Ren, L. Du, L. Carin, and D. Dunson. Logistic Stick-Breaking Process[J]. Machine Learning Research. 2011, 12(2): 203-239.

Google Scholar

[5] M. Zhou, L. Li, D. Dunson and L. Carin. Lognormal and Gamma Mixed Negative Binomial Regression[C]. /Proceedings of the 29th International Conference on Machine learning. Edinburgh, Scotland, UK: Spring Press, (2012).

Google Scholar