Text Mining of News Articles to Identify Security Threat Patterns in North East Nigeria

Automated extraction of threat intelligence from unstructured news text offers a scalable approach to monitoring security developments in conflict-affected regions of Nigeria. This study applied text mining techniques to 18,600 news articles published between 2019 and 2022 by six Nigerian and international news outlets covering security events in Borno, Yobe, Adamawa, and Gombe States. Articles were collected using web scraping scripts targeting freely accessible online publications. Named entity recognition was used to extract location, organisation, and actor mentions. Event classification using a custom keyword taxonomy distinguished between armed attacks, abductions, displacement events, and peace initiatives. Temporal analysis tracked monthly event frequencies by category. Results identified 3,847 distinct security-relevant events, with armed attacks constituting 44.1 percent, abductions 27.3 percent, and displacement events 18.6 percent. Borno State accounted for 61.2 percent of all classified events. Seasonal clustering showed increased attack frequency in November and December, potentially linked to dry season mobility. Cross-entity network analysis revealed frequent co-mention of specific local government areas and non-state actor identifiers. The study demonstrates that text mining of publicly available news data can generate structured threat intelligence to complement official security statistics. Limitations include publication bias in media coverage of rural incidents. Recommendations include developing an automated pipeline for continuous monitoring, validating text-derived event counts against ACLED data, and deploying the tool for academic and civil society conflict research in North East Nigeria.

Keywords: text mining, security threats, North East Nigeria, named entity recognition, conflict analytics