A decision-theoretic approach for segmental classification

DOI10.1214/13-AOAS657zbMATH Open1454.62206arXiv1007.4532OpenAlexW2048381900MaRDI QIDQ386773FDOQ386773

Authors: Christopher Yau, Chris Holmes 

Publication date: 10 December 2013

Published in: The Annals of Applied Statistics (Search for Journal in Brave)

Abstract: This paper is concerned with statistical methods for the segmental classification of linear sequence data where the task is to segment and classify the data according to an underlying hidden discrete state sequence. Such analysis is commonplace in the empirical sciences including genomics, finance and speech processing. In particular, we are interested in answering the following question: given data $y$ and a statistical model $pi(x,y)$ of the hidden states $x$, what should we report as the prediction $hatx$ under the posterior distribution $pi(x|y)$? That is, how should you make a prediction of the underlying states? We demonstrate that traditional approaches such as reporting the most probable state sequence or most probable set of marginal predictions can give undesirable classification artefacts and offer limited control over the properties of the prediction. We propose a decision theoretic approach using a novel class of Markov loss functions and report $hatx$ via the principle of minimum expected loss (maximum expected utility). We demonstrate that the sequence of minimum expected loss under the Markov loss function can be enumerated exactly using dynamic programming methods and that it offers flexibility and performance improvements over existing techniques. The result is generic and applicable to any probabilistic model on a sequence, such as Hidden Markov models, change point or product partition models.

Full work available at URL: https://arxiv.org/abs/1007.4532

Recommendations

• Bayesian classification of hidden Markov models
• A Bayesian approach for classification of Markov sources
• Time-series segmentation: A model and a method
• Sequence classification via large margin hidden Markov models
• Detecting Homogeneous Segments in DNA Sequences by Using Hidden Markov Models

zbMATH Keywords

Bayesiandecision theorysegmental classification

Mathematics Subject Classification ID

Classification and discrimination; cluster analysis (statistical aspects) (62H30) Markov processes: estimation; hidden Markov models (62M05) Bayesian problems; characterization of Bayes procedures (62C10)

Cites Work

• Product partition models for change point problems
• Statistical decision theory and Bayesian analysis. 2nd ed
• Title not available (Why is that?)
• Bayesian inference and state number determination for hidden Markov models: an application to the information content of the yield curve about inflation
• Analysis of default data using hidden Markov models
• Bridging Viterbi and posterior decoding: a generalized risk approach to hidden path inference based on hidden Markov models
• New Loss Functions in Bayesian Imaging
• Existence and uniqueness of countable one-dimensional Markov random fields
• Modelling Portfolio Defaults Using Hidden Markov Models with Covariates
• Reconstructing DNA copy number by penalized estimation and imputation

Cited In (9)

• Pairwise Markov models and hybrid segmentation approach
• Generalized species sampling priors with latent beta reinforcements
• Statistical testing of segment homogeneity in classification of piecewise-regular objects
• Nonparametric product partition models for multiple change-points analysis
• Estimation of Viterbi path in Bayesian hidden Markov models
• Regenerativity of Viterbi process for pairwise Markov models
• Exact Decoding of a Sequentially Markov Coalescent Model in Genetics
• Existence of infinite Viterbi path for pairwise Markov models
• On the accuracy of the MAP inference in HMMs

Uses Software

• BioHMM
• PICNIC
• TigrScan
• GlimmerHMM

This page was built for publication: A decision-theoretic approach for segmental classification

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q386773)