Probabilistic Classification Methods

The literature on classification algorithms has highlighted various data classification methods. Of these methods, probabilistic classification methods are amongst the most commonly used data classification methods. This chapter introduces two probabilistic models for classification: the hidden Markov model (HMM), which is a probabilistic generative model, and the logistic regression model (LR), which is a probabilistic discriminative model. The main target of an HMM is to find the probability of the observation being in a specific state where the observation is a probabilistic function of the state. The chapter presents the HMM and different techniques for training this model and their application in machine fault diagnosis. It provides a description of the LR model and a generalised LR model, which goes with the name MLR or multiple logistic regression, and its applications in machine fault diagnosis. B.S. Yan and J.M. Lee proposed a LR-based prognostic method for online performance degradation assessment and failure mode classification.