PoachNet: Predicting Poaching Using an Ontology-Based Knowledge Graph.

Area of Science:

• Conservation Technology
• Artificial Intelligence in Ecology
• Wildlife Management

Background:

• Poaching presents a critical threat to biodiversity and ecosystems globally.
• Current poaching prediction tools struggle with data inconsistencies and spatiotemporal complexities.
• Translating predictive insights into effective conservation strategies remains a significant challenge.

Purpose of the Study:

• To introduce PoachNet, a novel predictive system for inferring wildlife poaching likelihood.
• To integrate deep learning with Semantic Web reasoning for enhanced poaching prediction.
• To address the spatiotemporal complexity and actionability gap in current conservation tools.

Main Methods:

• Utilized elephant GPS data structured within an ontology-based knowledge graph.
• Employed a sequential neural network for predicting future elephant movements.
• Integrated predicted geo-locations into the knowledge graph and applied Semantic Web Rule Language (SWRL) for poaching risk inference.

Main Results:

• The PoachNet system successfully integrates deep learning predictions with semantic reasoning.
• Poaching risk is inferred based on geo-location predictions and predefined poaching logic.
• The geo-location prediction model demonstrated superior performance compared to state-of-the-art approaches.

Conclusions:

• PoachNet offers an advanced, actionable approach to predicting poaching hotspots.
• The integration of Semantic Web technologies provides a robust framework for conservation intelligence.
• This system advances the development of intelligent tools for wildlife protection and anti-poaching efforts.