This code deals with Bayesian online detection in univariate time-series of changepoints, i.e. abrupt variations in the generative parameters of a data, and regimes, i.e. subintervals of data between two consecutive changepoints. The main assumptions are that regimes have different mean value, in each regime data are conditionally Gaussian, observations in different regimes are independent (product partition model), and the number of regimes is not prespecified. The three algorithms in the library differ in the assumptions on the data generating process in each regime: - The first one is based on the BOCPD (Bayesian Online Change-Point Detection) algorithm that was introduced in 2007 by Ryan Adams and David MacKay [1]. Data in each regime are assumed to be independent and identically distributed. - The second one, MBO, assumes that within each regime data are described by a Markovian AR(1) model [2]. - The third one, SD-BOCPD (Score-Driven Bayesian Online Change-Point Detection), assumes that the data is Markovian within the regimes with time-varying autocorrelation [2]. This is achieved with a Score-Driven approach, a flexible class of observation driven time-varying parameter models. The code of each algorithm takes as input a univariate time-series and provides as output (i) the list of the most likely change-points between regimes, and (ii) a matrix with the run length (i.e. the time since the last change-point) posterior probability distribution at each time. The last output can also be generated as a figure. References: [1] Adams, R.P., & MacKay, D.J. (2007). Bayesian Online Changepoint Detection. arXiv:0710.3742 [2] Tsaknaki, I.-Y, Lillo F., Mazzarisi, P., Online Learning of Order Flow and Market Impact with Bayesian Change-Point Detection Methods, arXiv:2307.02375 Quantitative Finance, (in press 2024)