A change-point detection and clustering method in the recurrent-event context

Abstract

Change-point detection in the context of recurrent-event is a valuable analysis tool for identification of the intensity rate changes. It has been an interesting topic in many fields, such as medical studies, travel safety analysis, etc. If subgroups exist, clustering can be incorporated into the change-point detection to improve the quality of the results. This paper develops a new algorithm named Recurrent-K-means to detect the change-points of the intensity rates, and identify clusters of objects with recurrent events. It also proposes a test-based method to perform heuristic search in determining the number of underlying clusters. In this study, the objects are assumed to fall in several clusters while the objects in the same cluster share identical change-points. The event count for an object is assumed to be a non-homogeneous Poisson process with a piecewise constant intensity function. The methodology estimates the change-point as well as the intensity rates before and after the change-point for each cluster. The methodology establishes a clustering analysis based on K-means algorithm but enhances the procedure to be model based. The simulation study shows that the methodology performs well in parameter estimation and determination of the number of clusters in different scenarios. The methodology is applied to the UK coal mining disaster data to show its possible role in shaping government regulations and improving coal industry safety.

A non-parametric Bayesian change-point method for recurrent events

“…Notice that we assume one change-point mainly for three reasons. Firstly, many existing literature on change-point detection assumed one unknown change-point such as Raftery and Akman [42], Frobish and Ebrahimi [12], Li et al [26,28] or chose one-change point model after model-selection such as Li et al [27]. Secondly, DPMM is relatively complex.…”

“…Li et al [26] developed a hierarchical Bayesian finite mixture model (BFMM) to cluster the teenagers by change-points and risk rates before and after the change-points. Li et al [28] developed a new algorithm named Recurrent-K-means to tackle the same problem. These traditional models require the specification of the number of components a priori.…”

A Selective Review on Information Criteria in Multiple Change Point Detection

“…Similarly, ref. [50] adopted hypothesis testing combined with parametric bootstrapping in their proposed method for recurrent-event change point analysis, aiming to determine the number of clusters of change points in the UK coal mining disaster data.…”

Railway Axle Early Fatigue Crack Detection through Condition Monitoring Techniques

“…This algorithm improves the computational cost of Lloyd's algorithm as well as the quality of the final solution, improving performance [48]. This algorithm has been used in the literature to detect the change point [49]. In the field of predictive maintenance applied to rotating machinery, this technique has also been used to detect the defect initiation point in bearings, as in the case of [50,51].…”