LukeBer/clustering_disaggregated_conflict_sequences_africa

Clustering Disaggregated Conflict Sequences in Africa Based on Complex Spatio-Temporal Data # Clustering Disaggregated Conflict Sequences in Africa This repository is a collaborative work by **Sarah Deubner** and **Luke-Liam Bergmeier**. In this repository we introduces **four approaches to cluster conflict events with the goal of building conflict event sequences**. For each approach a class was defined, allowing to build sequences over time with different indicees and metrices. Additionally each class includes plotting and summary options to describe the sequences. Further more, there are options to compare different seuqneces using the compare sequence cluster class. Going further than build a sequence this repro also give an idea of how groups of sequences with similar conflict development patterns can be created. In preparation different normalization opperations are offered. In that way, the sequences development can be represented by a vector with a fixed dimension. This is crucial to build a cluster over sequences. This repository is a collaborative work by Luke-Liam Bergmeier and Sarah Deubner. # Final Results Refer to the notebook `notebooks/final_sequence_clustering` to get an overview of all four sequence types their comparison, nomalization and clustering. # Dataset The analysis of conflict data in Africa is based on two datasets, which are aggregated: the Armed Conflict Location and Event Data Project (ACLED) and the Uppsala Conflict Data Program (UCDP). Both are event-based datasets, meaning that the unit of observation is an individual incident of political violence or disorder. Each event specifies the actors involved, the type of incident, the time and place of occurrence, as well as additional details such as fatalities and a note. We **do not redistribute** the datasets in this repository. You must **download them yourself** and accept the respective terms of use: - ACLED typically requires registration and acceptance of a user agreement. - UCDP data is available under its own terms; please review before use. The `data_loder.py …