Data-driven conflict classification exposes weak predictive indicators

Triangle of Madness This repository contains the code and the data that was generated to perform the analysis shown in our paper (Kushwaha et al., Royal Society Open Science, 2025). This repository can be used to reproduce the results (figures) shown in the paper. Instructions to reproduce results from the paper is shown in the next section. Reproducing figures shown in paper In order to reproduce the plots from the paper, follow the following instructions. Note: The following instructions work for linux systems with Conda preinstalled. 1) Extract the contents of this zip folder. 2) Then enter the folder: $ cd triangle_of_madness 3) Install the python package using the "initial_installation.sh" file: $ bash initial_installation.sh During this installation process, a new conda environment will be created in your system and you will be prompted to name this new environment. Use this name in the next step. 4) Once the installation is complete, activate the installed conda environment. $ conda activate <ENV NAME> 5) Run the "reproduce_plots.ipynb" jupyter notebook to reproduce results. Note: Figure 1 takes the most and significant amount of time to produce. Have some patience while running the notebook! Disclaimer The code provided here is just a snapshot of the codebase at the time of the publication of our article. For any future potential updates check the corresponsing Github page: https://github.com/NirajKushwaha/triangle_of_madness. The version of the code provided in this Zenodo repository is in the branch called RSOS_2025.