Armed conflict in Nigeria poses a persistent threat to national stability, driven by complex socio-economic, political, and environmental factors. This study introduces a novel hybrid approach that integrates Association Rule Mining (ARM) with Link Prediction (LP) to improve the prediction of the likelihood and severity of armed conflict outbreak factors across Nigeria. Utilizing data from the Armed Conflict Location & Event Data Project (ACLED) database (1997–2024), comprising over 187,000 conflict event records [27], and the Nigeria Visualized Platform, the research explores co-occurrence patterns and structural relationships among conflict-related variables, such as unemployment rates, fatality levels, actor types, and event locations. The Apriori algorithm was applied to extract 225 association rules from a transactional conflict dataset, which were then refined using link prediction (LP) metrics—including Common Neighbors, Jaccard Coefficient, and Adamic-Adar Index—within a bipartite graph structure. Results show that the hybrid analytical model significantly improves rule quality. Specifically, average lift increased from 2.97 to 4.21, average confidence rose from 48.5% to 61.2%, and median support improved from 2.4% to 3.1% compared to traditional ARM alone. The final set of 68 high-confidence rules included strong associations between very high unemployment and protests (lift = 3.85, confidence = 61%), and correlations between specific actors, such as Boko Haram, and high-fatality events (lift = 3.52, confidence = 63%). This study illustrates the empirical value of integrating data mining techniques with network-theoretical approaches to improve the detection of structural conflict precursors. Future research should incorporate temporal dynamics and geospatial modeling to improve the generalizability and policy relevance of the proposed framework.
