About

The "Patterns of Conflict" report identifies and compares conflict patterns across various countries. This process involves aggregating historical conflict data and matching similar patterns of conflict-related events. The methodology focuses on identifying trends and potential future scenarios based on historical data. The objective is to provide a predictive insight into how conflict patterns may evolve, aiding in better-informed strategic planning and decision-making.

The methodology in the "Patterns of Conflict" report is centered on a comparative analysis of conflict-related data across countries. It involves the following steps:

1. Data collection. The data used in the "Patterns of Conflict" report is sourced from the Uppsala Conflict Data Program (UCDP), a comprehensive database that records and codes data on conflict and associated events worldwide. Specifically, the report makes use of the "best" estimate variable for battle-related deaths provided by UCDP (see https://ucdp.uu.se/downloads/brd/ucdp-brd-codebook.pdf)

2. Short sequences of casualty data are compared to each other using various algorithms (DTW, Euclidean distance), which allow us to identify similar shapes in the data, even ones that may be out of sync temporally. A distance threshold is applied to select only sequences that are close matches.

3. The model then predicts potential increases or decreases in conflict-related fatalities based on an average of past patterns.

• Conflict-Fatalities:UCDP Dataset, aggregated at the country-monthly level.
• UCDP Georeferenced Event Dataset.
• UCDP Candidate Events Dataset (to get the latest data).

The applied model operates by examining recent events within a country and aligning them with historical occurrences. It discerns patterns in the temporal evolution of incidents, enabling the identification of analogous scenarios from the past. This matching process contributes to a comprehensive understanding of when and where comparable situations have historically manifested. Consequently, the model plays a pivotal role in predicting the future trajectory of potential conflict-related scenarios based on these historical parallels, called ‘Past Future’.

To identify match in historical sequences, we employ dynamic time warping (DTW) distance. In contrast to the Euclidean distance, DTW offers greater flexibility in accommodating variations in time and window length. DTW works by aligning the two sequences in a way that minimizes the total distance between corresponding points, allowing for both temporal shifts and local deformations. This alignment is achieved by warping the time axis of one sequence with respect to the other. The warping path represents the optimal alignment, and the DTW distance is the cumulative sum of the distances along this path. One of the key advantages of DTW is its ability to handle sequences of unequal length and to flexibly adapt to local variations in timing. The DTW distance is computed, and if it falls below a predefined threshold, the historical sequence is classified as a match.