This library performs regime detection in the aggregated order flow time-series and market impact analysis. The required input file is in the format of the message file of the LOBSTER database [1] containing all the events in a Limit Order Book market for trading day. The code is also able to parse multiple trading days, i.e. multiple message files. For each message file, the aggregated order flow (volume of buy market orders minus volume of sell market orders) on a user-specified number of trades is created. Then, the SD-BOCPD (Score-Driven Bayesian Online Change-Point Detection)[2] companion code is used to obtain a list of change-points between regimes. The list of change-points between regimes is used to provide the average market impact (signed price change) within a regime as a function of time or of volume. Additionally, the code provides online predictions of order flow and market impact [3]. References: [1] https://lobsterdata.com/info/DataStructure.php [2] https://sobigdata.d4science.org/catalogue-sobigdata?path=/dataset/score-driven_bayesian_online_change_point_detection_sd-bocpd_ [3] 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)