Abstract:
|
In this paper, we consider the problem of modeling a matrix of count data, where multiple features are observed as counts over a number of samples. Due to the nature of the data generating mechanism, such data are often characterized by a high number of zeros and overdispersion. In order to take into account the skewness and heterogeneity of the data, some type of normalization and regularization is necessary for conducting inference on the occurrences of features across samples. We propose a zero-inflated Poisson mixture modeling framework that incorporates a model-based normalization through prior distributions with mean constraints, as well as a feature selection mechanism, which allows us to identify a parsimonious set of discriminatory features, while simultaneously clustering the samples into homogenous groups. Using simulated data, we show how our approach improves on the accuracy of the clustering with respect to more standard approaches for the analysis of count data. We also present an application to a bag-of-words benchmark data set, where the features are represented by the frequencies of occurrence of each word.
|