How AI can help prevent elephant poaching

A new AI system could help prevent elephant poaching in Malaysia, according to researchers at Cardiff University.

PoachNet is a new machine-learning tool designed by integrating deep learning, elephant GPS data, and elephant behavioral knowledge, to help predict and prevent poaching in Sabah.

Cardiff University research from the School of Computer Science and Informatics and School of Biosciences used machine-learning and a smart database to create a system that can predict poaching risks. The research, "PoachNet: Predicting Poaching Using an Ontology-Based Knowledge Graph," was published in Sensors.

PoachNet has been developed to help predict future elephant geo-locations and uses this information to identify poaching risks based on proximity to identified hazardous areas.

Naeima Hamed, doctoral researcher at Cardiff University's School of Computer Science and Informatics, said, "Elephant GPS data is analyzed with a special type of AI—a sequential neural network—to predict their movements. These predictions are added to the knowledge graph in a meaningful way—then PoachNet uses a rule-based system to apply poaching rules and detect hidden patterns in the data.

"We found that, when tested, PoachNet was more accurate than other leading methods, consistently performing better. By handling the complexity of time and space data and turning predictions into practical rules, PoachNet offers a big improvement in tracking and protecting elephants."

Previous approaches to preventing elephant poaching have focused on specific aspects, such as social network analysis, multimedia data mining, or models based on ranger patrol data—but Cardiff University's PoachNet approach incorporates understanding of wildlife dynamics.

The researchers hope that PoachNet can assist in informing strategies for tackling anti-poaching in Sabah, by helping with resource allocation based on its predictions. It can also guide the deployment of motion-activated camera traps in areas most likely to have anticipated poaching crimes.

Professor Omer Rana, International Dean for the Middle East and Professor of Performance Engineering at Cardiff University School of Computer Science and Informatics, said, "PoachNet is a unique software tool that integrates semantically modeled regional data sources with emerging machine learning algorithms and semantic reasoning. PoachNet addresses a challenge that continues to affect communities supporting endangered species.

"Both climate change and economics are leading to significant impact across this interface between human activity and natural habitats. The data-driven approach adopted in PoachNet can also be generalized to support other similar localities and national parks—enabling more efficient use of law enforcement and government resources."

Professor Benoit Goossens, director of the Danau Girang Field Centre and Professor at Cardiff University's School of Biosciences, said, "Habitat loss, human-elephant conflict, and poaching threaten Bornean elephants. Despite global anti-poaching efforts, the illegal ivory trade continues to drive poaching, reducing the population to fewer than 1500. We hope that PoachNet can assist in poaching prevention methods, therefore helping to ensure the safety of the elephant population in Sabah for the future."

The researchers also hope that PoachNet can be expanded by integrating additional wildlife data sources—such as acoustic sensors to detect gunshots or vehicle noise, satellite imagery to monitor habitat changes and human activity, and crime intelligence.

More information: Naeima Hamed et al, PoachNet: Predicting Poaching Using an Ontology-Based Knowledge Graph, Sensors (2024). DOI: 10.3390/s24248142