Using AI to predict livestock growth and protect sensitive data

6 August 2026

Dr Shoujin Wang from the UTS Data Science Institute (DSI) is developing innovative AI solutions that could help improve productivity and sustainability across Australia's agricultural sector whilst ensuring farmers retain control of their data.

Dr Shoujin Wang speaking at Research Cafe

Speaking about his latest research, Shoujin explained that accurate livestock growth prediction is essential for helping farmers optimise farm management and improve outcomes.

However, developing effective AI models can be challenging because farmers are often reluctant to share management data that may contain commercially sensitive information.

“This makes it very difficult to build a model to centralise all the data for growth prediction,” Shoujin said.

Dr Victor Chu, Leader of A-Theme at UTS DSI, also at the venue, said many farmers are operating under data-poor environments where critical, but business sensitive data are not easily sharable in the industry. This limits the ability of AI to deliver on its potential.

“To unlock this barrier, our data scientists have developed a collaborative framework that allows multiple farmers to jointly train a machine learning model without ever having to share their private and sensitive business data.”

Each farm trains a local model using its own data and shares only model parameters which are then combined into a global model. This approach enables farms to benefit from insights gathered across the industry while keeping sensitive information securely on site.

The team has also enhanced the system by developing personalised models that combine broad industry knowledge with farm-specific characteristics. This allows the AI system to provide more accurate and tailored livestock growth predictions for individual farms.

To unlock this barrier, our data scientists have developed a collaborative framework that allows multiple farmers to jointly train a machine learning model without ever having to share their private and sensitive business data.

Testing on real-world farm data demonstrated that the approach improved prediction accuracy while maintaining strong privacy protections.

“Our research highlights how building trust with end users, respecting data ownership and protecting privacy can support the development of practical, trustworthy AI solutions that deliver real value to Australia's agricultural sector,” Shoujin concluded.