Multivariate count time series segmentation with "sums and shares" and Poisson lognormal mixture models: a comparative study using pedestrian flows within a multimodal transport hub.
Paul de NaillyEtienne CômeLatifa OukhellouAllou SaméJacques FerriereYasmine Merad-BoudiaPublished in: Adv. Data Anal. Classif. (2024)
Keyphrases
- mixture model
- finite mixtures
- gaussian mixture model
- em algorithm
- generative model
- density estimation
- multivariate time series
- bayesian information criterion
- probabilistic model
- expectation maximization
- unsupervised learning
- level set
- multivariate data
- finite mixture models
- maximum likelihood
- segmentation algorithm
- segmentation method
- language model
- model selection
- mixture modeling
- regression model
- machine learning
- probability density function
- background subtraction
- random variables
- image segmentation
- unsupervised image segmentation
- automatic model selection
- generalized em algorithm
- dirichlet distribution
- model based clustering
- unsupervised segmentation
- pixel level
- supervised learning
- pattern recognition
- computer vision
- information retrieval