Mixture Model Classifier Research Articles

Modelling Count Responses with Overdispersion

We frequently encounter outcomes of count that have extra variation. This paper considers several alternative models for overdispersed count responses such as a quasi-Poisson model, zero-inflated Poisson model and a negative binomial model with a special focus on a generalized linear mixed model. We also explain various goodness-of-fit criteria by discussing their appropriateness of applicability and cautions on misuses according to the patterns of response categories. The overdispersion models for counts data have been explained through two examples with different response patterns. Poisson distribution is the standard model for count responses, for which the variance of responses is expected to equal mean. But violations in the assumptions characterizing a Poisson distribution may lead to the presence of extra variation in the analysis of counts and rates in longitudinal studies. There may exist a dependence between the elemental units in a study where experimental units are clusters. When responses are observed in clusters they frequently exhibit extra variation than the permitted variance of the assumed model. We often encounter this extra variation in a real data of count or binomial responses. Overdispersion may sometimes be observed when the residual variation obtained is greater than which can be attributed to the sampling variation assumed by the model. McCullagh and Nelder (1989) showed that overdispersion is not uncommon in practice. Overdis- persion should be considered deliberately in modelling count responses. The ordinary Poisson general linear model(GLM) cannot be fitted well in the presence of overdispersion. In the analysis of clus- tered or longitudinal data the generalized linear mixed model(GLMM) and the generalized estimating equations(GEE) approaches are most popular. The GLMM incorporates random subject effects into a GLM by allowing subjects variability; however, the GEE method solves equations that include the correlation structure of repeated responses to estimate regression coefficients. Thall and Vail (1990) discussed covariance models for longitudinal count data with overdisper- sion, and Jowaheer and Sutradhar (2002) applied a GEE method to analyze longitudinal count data; however, Sutradhar et al. (2007) suggested three kinds of mixture models to account for the overdis- persion in binomial data. The beta-binomial, the finite mixture, and the zero-inflated binomial model all belong to the same class of mixture models. They suggested a chi-squared type goodness-of-fit statistic to test the assumed null model against the alternative three kinds of models without con- sidering covariate variables. Recently, Morel and Neerchal (2012) have extensively studied various overdispersion models using SAS.

Multi-appliance recognition system with hybrid SVM/GMM classifier in ubiquitous smart home

Ubiquitous computing provides convenient and fast information distribution service by using sensor nodes and wireless network, and a good household appliance recognition system will allow users to effectively understand the household appliance usage and develop habits of power preservation. At present, smart meters convert the information of traditional electric meters to easily accessible digital information, based on which, the household appliance recognition service can be carried out. However, it is different from video or audio recognition service, when a variety of electrical appliances run, they will all have individual impact on power consumption, thereby resulting in the difficulties in recognition. Presently, the complex current information arising from many household appliances also increases the difficulty in extracting power features. For addressing the challenge, this study proposes a set of multi-appliance recognition system, which designs a single smart meter using a current sensor and a voltage sensor in combination with a microprocessor to meter multi-appliances. After fuzzy processing of the power information are read through the smart meter and extraction of the power features, electric appliances are classified using the hybrid Support Vector Machine/Gaussian Mixture Model (SVM/GMM) classification model. GMM is mainly used describe the wave distribution situation according to the current information, so as to find the power similarity; while SVM is used to classify the power features of different electric appliances, so as to summarize the classification properties of different electric appliances and establish a classification model. Finally, the household appliances that are in use can be recognized with the household power supply terminal, and their information can be reported to users through wired or wireless network to achieve ubiquitous recognition service. This study has developed and implemented this system prototype, and is used to prove its design theory.

Permutation Inference for a Class of Mixture Models

In statistical surveys, people are often asked to express evaluations on several topics or to make an ordered arrangement in a list of objects (items, services, sentences, etc.); thus, the analysis of ratings and rankings is receiving a growing interest in many fields. In this framework, we develop a testing procedure for a class of mixture models with covariates (defined as CUB models), proposed by Piccolo (2003) and D'Elia and Piccolo (2005) and generally developed in a parametric context. Instead, we propose a nonparametric solution to perform inference on CUB models, specifically on the coefficients of the covariates. A simulation study proves that this approach is more appropriate in some specific data settings, mostly for small sample sizes.

Mixture models with an improper component

A class of mixture models in which a component is associated with an improper distribution is introduced. This component is intended mainly for outliers. The models are motivated by the EM algorithm, and are fitted by its simple adaptation. They are illustrated on several examples with large samples, one of them about transactions of residential properties in Wellington, New Zealand, in 2006.

A Real-Time System for Pedestrian Detection Based on GMM and HOG

The detection of pedestrian which has been widely used in digital surveillance systems is a popular topic in computer vision. This paper mainly discusses a system of pedestrian detection in video sequences captured from a stationary camera hanging in a public scene. We describe an efficient system combining background subtraction based on Gaussian Mixture Model (GMM) and object classification based on Histograms of Oriented Gradients (HOG). We first process moving objects segmentation using GMM. Then a HOG detector is used to classify the moving objects into person and none-person. Experimental results on video sequences have demonstrated that the real-time tracking system can process 15 to 30 frames per second robustly with a high accuracy.

Conditional exact tests for Markovianity and reversibility in multiple categorical sequences

We propose conditional exact tests based on sufficient statistics to compare exchangeability, Markov exchangeability, and Markov exchangeability of the reversible type, when data consist of several sequences of categorical data. As particular cases, we can compare three classes of mixture models (mixtures of i.i.d. sequences, mixtures of Markov chains and of reversible Markov chains) and we can test Markovianity and reversibility of a single sequence.

Finite mixture models for clustering multilevel data with multiple cluster structures

Finite mixture models are useful tools for clustering two-way datasets within a sound statistical framework which can assess some important questions, such as how many clusters are there in the data. Models that can also be used for clustering multilevel data have been proposed, with the intent to produce clusterings of units at every level on the basis of all the available variables, considering the hierarchical structure of the dataset. This paper introduces a new class of mixture models for datasets with two levels that makes it possible to discover a clustering of level 2 units and different clusterings of level 1 units corresponding to different subsets of the variables (multiple cluster structures). This new class is obtained by adapting a mixture model proposed to identify multiple cluster structures in a data matrix to the multilevel situation. The usefulness of the new method is shown using simulated data and a real example.

Abnormal audio recognition algorithm based on MFCC and short-term energy

Concerning the high complexity and low rate in abnormal audio recognition,the abnormal audio recognition system based on the Mel-Frequency Cepstrum Coefficients(MFCC)and short-term energy was proposed.This feature vector made the Gaussian Mixture Model(GMM)classifier outperform MFCC and Differential MFCC features in classification.The classifier can achieve an average recognition rate of more than 90%,and small computational complexity.The steps of system implementation were elaborated.The simulation results prove the effectiveness of the proposed algorithm.

Modeling Total Solar Irradiance Variations Using Automated Classification Software on Mount Wilson Data

We present results using the AutoClass analysis application available at NASA/Ames Intelligent Systems Div. (2002) which is a Bayesian, finite mixture model classification system developed by Cheeseman and Stutz (1996). We apply this system to Mount Wilson Solar Observatory (MWO) intensity and magnetogram images and classify individual pixels on the solar surface to calculate daily indices that are then correlated with total solar irradiance (TSI) to yield a set of regression coefficients. This approach allows us to model the TSI with a correlation of better than 0.96 for the period 1996 to 2007. These regression coefficients applied to classified pixels on the observed solar surface allow the construction of images of the Sun as it would be seen by TSI measuring instruments like the Solar Bolometric Imager recently flown by Foukal et al., (2004). As a consequence of the very high correlation we achieve in reproducing the TSI record, our approach holds out the possibility of creating an on-going, accurate, independent estimate of TSI variations fromground-based observations which could be used to compare, and identify the sources of disagreement among, TSI observations from the various satellite instruments and to fill in gaps in the satellite record. Further, our spatially-resolved images should assist in characterizing the particular solar surface regions associated with TSI variations. Also, since the particular set of MWO data on which this analysis is based is available on a daily basis back to at least 1985, and on an intermittent basis before then, it will be possible to estimate the TSI emission due to identified solar surface features at several solar minima to constrain the role surface magnetic effects have on long-term trends in solar energy output.

Multivariate Weibull Mixtures with Proportional Hazard Restrictions for Dwell-Time-Based Session Clustering with Incomplete Data

Summary Emanating from classical Weibull mixture models we propose a framework for clustering survival data with various more parsimonious models by imposing restrictions on the distributional parameters. We show that these restrictions on the Weibull mixtures correspond to different proportional hazard restrictions across mixture components and Web page areas. A parametric cluster approach based on the EM algorithm is carried out on a multivariate data set. Our model set-up encompasses incomplete-data structures as well as censoring observations. We apply the methodology on retail data stemming from a global e-commerce company. Sessions are clustered with respect to the dwell times that a user spends on certain page areas. The cluster solution that is found allows for a detailed examination of the navigation behaviour in terms of the hazard and survivor functions within each component.

Finding Relevant Linear Manifolds in Classification by Gaussian Mixtures

In this article, we present a strategy for producing low-dimensional projections that maximally separate the classes in Gaussian Mixture Model classification. The most revealing linear manifolds are those along which the classes are maximally separable. Here we consider a particular probability product kernel as a measure of similarity or affinity between the class-conditional distributions. It takes an appealing closed analytical form in the case of Gaussian mixture components. The performance of the proposed strategy has been evaluated on real data.

Speaker Identification Using Instantaneous Frequencies

This paper presents an experimental evaluation of different features for use in speaker identification. The features are tested using speech data provided by the chains corpus, in a closed-set speaker identification task. The main objective of the paper is to present a novel parametrization of speech that is based on the AM-FM representation of the speech signal and to assess the utility of these features in the context of speaker identification. In order to explore the extent to which different instantaneous frequencies due to the presence of formants and harmonics in the speech signal may predict a speaker's identity, this work evaluates three different decompositions of the speech signal within the same AM-FM framework: a first setup has been used previously for formant tracking, a second setup is designed to enhance familiar resonances below 4000 Hz, and a third setup is designed to approximate the bandwidth scaling of the filters conventionally used in the extraction of Mel-fequency cepstral coefficients (MFCCs). From each of the proposed setups, parameters are extracted and used in a closed-set text-independent speaker identification task. The performance of the new featural representation is compared with results obtained adopting MFCC and RASTA-PLP features in the context of a generic Gaussian mixture model (GMM) classification system. In evaluating the novel features, we look selectively at information for speaker identification contained in the frequency range 0-4000 Hz and 4000-8000 Hz, as the instantaneous frequencies revealed by the AM-FM approach suggest the presence of structures not well known from conventional spectrographic analyses. Accuracy results obtained using the new parametrization perform as well as conventional MFCC parameters within the same reference system, when tested and trained on modally voiced speech which is mismatched in both channel and style. When the testing material is whispered speech, the new parameters provide better results than any of the other features tested, although they remain far from ideal in this limiting case.

AUTOMATED IDENTIFICATION OF EYE DISEASES USING HIGHER-ORDER SPECTRA

A computer-based intelligent system for the classification of eye diseases can be very useful for their diagnosis and management. With age, the incidence of ocular pathology rises, thereby decreasing normal eye function. The most common causes of age-related eye disorders and visual impairment in the elderly are cataracts and iridocyclitis (inflammation of the iris, i.e. the colored part of the eye, and of the ciliary body). For proper care and management of eyes, we need a system which can automatically classify these eye diseases. The method proposed in this study is based on higher-order spectral (HOS) features that capture contour and shape information, while providing robustness to shift, rotation, changes in size, and noise. The parameters are extracted from the raw images using the HOS techniques, and fed to the classifiers for classification. This paper presents the classification of three kinds of eye classes using four-layer feedforward and Gaussian mixture model (GMM) classifiers. Our protocol used 122 subjects who had three different kinds of eye disease conditions. We demonstrated a sensitivity of 100% for the classifier, with a specificity of 90%. Our systems are clinically ready to test on large data sets.

Learning correlations using the mixture-of-subsets model

Using a mixture of random variables to model data is a tried-and-tested method common in data mining, machine learning, and statistics. By using mixture modeling it is often possible to accurately model even complex, multimodal data via very simple components. However, the classical mixture model assumes that a data point is generated by a single component in the model. A lot of datasets can be modeled closer to the underlying reality if we drop this restriction. We propose a probabilistic framework, the mixture-of-subsets (MOS) model , by making two fundamental changes to the classical mixture model. First, we allow a data point to be generated by a set of components, rather than just a single component. Next, we limit the number of data attributes that each component can influence. We also propose an EM framework to learn the MOS model from a dataset, and experimentally evaluate it on real, high-dimensional datasets. Our results show that the MOS model learned from the data represents the underlying nature of the data accurately.

Flexible Cure Rate Modeling Under Latent Activation Schemes

With rapid improvements in medical treatment and health care, many datasets dealing with time to relapse or death now reveal a substantial portion of patients who are cured (i.e., who never experience the event). Extended survival models called cure rate models account for the probability of a subject being cured and can be broadly classified into the classical mixture models of Berkson and Gage (BG type) or the stochastic tumor models pioneered by Yakovlev and extended to a hierarchical framework by Chen, Ibrahim, and Sinha (YCIS type). Recent developments in Bayesian hierarchical cure models have evoked significant interest regarding relationships and preferences between these two classes of models. Our present work proposes a unifying class of cure rate models that facilitates flexible hierarchical model building while including both existing cure model classes as special cases. This unifying class enables robust modeling by accounting for uncertainty in underlying mechanisms leading to cure. Issues such as regressing on the cure fraction and propriety of the associated posterior distributions under different modeling assumptions are also discussed. Finally, we offer a simulation study and also illustrate with two datasets (on melanoma and breast cancer) that reveal our framework's ability to distinguish among underlying mechanisms that lead to relapse and cure.

Gaussian mixture model classification of odontocetes in the Southern California Bight and the Gulf of California

A method for the automatic classification of free-ranging delphinid vocalizations is presented. The vocalizations of short-beaked and long-beaked common (Delphinus delphis and Delphinus capensis), Pacific white-sided (Lagenorhynchus obliquidens), and bottlenose (Tursiops truncatus) dolphins were recorded in a pelagic environment of the Southern California Bight and the Gulf of California over a period of 4 years. Cepstral feature vectors are extracted from call data which contain simultaneous overlapping whistles, burst-pulses, and clicks from a single species. These features are grouped into multisecond segments. A portion of the data is used to train Gaussian mixture models of varying orders for each species. The remaining call data are used to test the performance of the models. Species are predicted based upon probabilistic measures of model similarity with test segment groups having durations between 1 and 25 s. For this data set, 256 mixture Gaussian mixture models and segments of at least 10 s of call data resulted in the best classification results. The classifier predicts the species of groups with 67%-75% accuracy depending upon the partitioning of the training and test data.

Gaussian mixture model classification: A projection pursuit approach

Gaussian mixture models (GMM) are commonly employed in nonparametric supervised classification. In high-dimensional problems it is often the case that information relevant to the separation of the classes is contained in a few directions. A GMM fitting procedure oriented to supervised classification is proposed, with the aim of reducing the number of free parameters. It resorts to projection pursuit as a dimension reduction method and combines it with GM modelling of class-conditional densities. In its derivation, issues regarding the forward and backward projection pursuit algorithms are discussed. The proposed procedure avoids the “curse of dimensionality”, is able to model structure in subspaces and regularizes the classification model. Its performance is illustrated on a simulation experiment and on a real data set, in comparison with other GMM-based classification methods.

Mixture Models and Subpopulation Classification: A Pharmacokinetic Simulation Study and Application to Metoprolol CYP2D6 Phenotype

Mixture models are applied in population pharmacometrics to characterize underlying population distributions that are not adequately approximated by a single normal or lognormal distribution. In addition to obtaining individualized maximum a posteriori Bayesian post hoc parameter estimates, the subpopulation to which an individual was classified can be determined. However, the accuracy of the classification of subjects to subpopulations is not well studied. We investigated the impact of several factors on the accuracy of classification in mixture models applied to pharmacokinetics using a simulation strategy. The availability of actual subject data allowed us to evaluate mixture model classification in a potentially common application, namely, the classification of clearance into poor metabolizer (PM) or extensive metabolizer (EM) subgroups with the known phenotype status in subjects receiving metoprolol. The factors explored in the simulation study were the magnitude of difference between the clearances in two subpopulations, the between subject variability in clearance, the mixing-fraction, and the population sample size. Populations were simulated at various levels of the above factors and analyzed with a mixture model using NONMEM. The population pharmacokinetics of metoprolol were modeled with the EM/PM phenotype as a known covariate, and without the phenotype covariate using a mixture model. Within the range of scenarios studied, the proportion of subjects classified into the correct subpopulation was high. The simulation-estimation study suggests that a greater separation between two subpopulations, a smaller variability in the parameter distribution, a larger sample size, and a smaller size subpopulation tend to be associated with a greater accuracy of subpopulation classification when a mixture model is applied to pharmacokinetic data. In a population pharmacokinetic analysis of metoprolol, a drug that undergoes polymorphic metabolism, it was possible to correctly identify phenotype status using a mixture model.

On Identifiability in Capture–Recapture Models

We study the issue of identifiability of mixture models in the context of capture-recapture abundance estimation for closed populations. Such models are used to take account of individual heterogeneity in capture probabilities, but their validity was recently questioned by Link (2003, Biometrics 59, 1123-1130) on the basis of their nonidentifiability. We give a general criterion for identifiability of the mixing distribution, and apply it to establish identifiability within families of mixing distributions that are commonly used in this context, including finite and beta mixtures. Our analysis covers binomial and geometrically distributed outcomes. In an example we highlight the difference between the identifiability issue considered here and that in classical binomial mixture models.

Assessing Heterogeneity in Discrete Choice Models Using a Dirichlet Process Prior

The finite normal mixture model has emerged as a dominant methodology for assessing heterogeneity in choice models. Although it extends the classic mixture models by allowing within component variablility, it requires that a relatively large number of models be separately estimated and fairly difficult test procedures to determine the “correct” number of mixing components. We present a very general formulation, based on Dirichlet Process Piror, which yields the number and composition of mixing components a posteriori, obviating the need for post hoc test procedures and is capable of approximating any target heterogeneity distribution. Adapting Stephens’ (2000) algorithm allows the determination of ‘substantively’ different clusters, as well as a way to sidestep problems arising from label-switching and overlapping mixtures. These methods are illustrated both on simulated data and A.C. Nielsen scanner panel data for liquid detergents. We find that the large number of mixing components required to adequately represent the heterogeneity distribution can be reduced in practice to a far smaller number of segments of managerial relevance.