Breakpoint: an R package to detect multiple change-points in continuous and count data via the croS-en-tropy method

The Cross-Entropy (CE) method is an evolutionary stochastic optimization method that was originally developed to estimate probabilities of rare events. Later with the advances of the CE method, it could be used to solve dicult combinatorial optimization problems (COPs) as well. A change-point problem may fall under the broader category of COPs but it may also be considered within control theory. We de ne the change-point (or break-point) problem as the process of detecting the location where a distributional change has occurred in an observed sequence of data. Thus, it is a mixture of both estimation and optimization problems. In real life scenarios, there can be more than one change-point in a process. Therefore, a successful procedure should be able to detect both the the number of change-points and their locations simultaneously with a high level of sensitivity. The breakpoint" R package, which is available on CRAN, can be used to solve these two types of problems. It uses the CE method in the optimization step to obtain the locations of the change-points. An information criterion is used to obtain the number of change-points. The calculations can be carried out both sequentially and in parallel with the use of multiple-cores. In this study, we assess the di erent functionalities of the package with multiple data sources. At rst, a comprehensive simulation study was carried out to detect both the number and the locations of change-points in continuous as well as count data. The results are compared with some of the popular methods that are available in the literature. Finally, we applied the procedures to real data originated from di erent scienti c streams. Our results suggest that the proposed methods in the breakpoint" package are e ective in detecting both the number and locations of change-points in continuous and count data.