Supplementary Data and Code: Change Point Analysis to decode Economic Crisis Information

<p>Raw data, results and Python code of the corresponding publication &quot;Efficient Multi-Change Point Analysis to decode Economic Crisis Information from the S&amp;P500 Mean Market Correlation&quot; (accepted in: Entropy; Section: Complexity; Special Issue: Complexity in Finance). The change point analysis can be performed using the <a href="https://anticpy.readthedocs.io/en/latest/">documented</a> Python package <a href="https://github.com/MartinHessler/antiCPy"><em>antiCPy</em></a>. Some further helpful Python scripts are provided here under a <em>GNU General Public License v3.0.</em></p> <p>In Data_Generation you can find a list of</p> <ol> <li>S&amp;P500 companies which are considered in the analysis,</li> <li>a jupyter notebook to create the correlation time series.</li> </ol> <p>Data_Preprocessing contains the</p> <ol> <li>S&amp;P500 mean market correlation Financial_Time_Series_Centered_Interval__42days.csv,</li> <li>the Python code to thin it,</li> <li>the thinned data saved as .npy file,</li> <li>the time scale is saved <ul> <li>as integer numbers in thinned_time_thinning40.npy,</li> <li>as datetime in TimeScale_FinancialData.npy.</li> </ul> </li> </ol> <p>In Change_Point_Analysis you find the following files:</p> <ol> <li>cp_probs...npy contain the joint probabilities of the corresponding change point configurations (The joint probabilities are saved corresponding to the order in which Python&#39;s <em>itertools.combinations() </em>creates the configurations. This holds also for the cp_probs_5_cps.npy for which the whole combinations array is not saved for memory reasons),</li> <li>cp_pdfs...npy contain the marginal probability density functions of the ordinal change point positions averaged over all configurations,</li> <li>cp_configs...npy contain the configurations,</li> <li>segment_fit...npy contain the segment fit data,</li> <li>segment_fit_variance...npy contain corresponding variances,</li> <li>In the case of five change points only the plotted 1st, 13th and 26th most probable configuration in config_ranking_CP1_5.npy for memory reasons.</li> <li>the data and results of figure 3 for each crisis event can be found in the corresponding directory&#39;s folders: <ul> <li>blue corresponds to the pre-crisis data segments,</li> <li>green corresponds to the data segments up to the green vertical dotted line,</li> <li>red corresponds to the longest data segments incorporating near and in-crisis data.</li> </ul> </li> </ol> <p>&nbsp;</p> <p>&nbsp;</p>