下载PDF

Amazon Web Services > 实例探究 > From Cows to the Cloud: How TINE is Revolutionizing the Norwegian Dairy Industry Using Machine Learning on AWS

From Cows to the Cloud: How TINE is Revolutionizing the Norwegian Dairy Industry Using Machine Learning on AWS

技术

平台即服务 (PaaS) - 数据管理平台

适用行业

农业
食品与饮料

适用功能

离散制造
质量保证

用例

预测性维护
机器状态监测
实时定位系统 (RTLS)

服务

数据科学服务
云规划/设计/实施服务

挑战

TINE SA, a Norwegian cooperative owned by farmers, has been collaborating with Norway’s farmers for over 160 years to understand their challenges and help them drive efficiency, productivity, and high-quality dairy product development. As market demands increase, so does each farmer’s need to bring products to market more efficiently. TINE has a long history of developing decision-making tools for farmers and collecting data from different dairy products at the farm level. However, as TINE considered the future of dairy production at both the farm and national level, its data science team realized there would be changes to the breadth and depth of its data sources, types of data, and data analysis capabilities available to develop decision-making tools for farmers to use in production. TINE knew it would have to change its approach to data and technology by becoming more data driven as an organization to drive better predictability of milk production and other key data points related to a cow’s health and the quality of milk produced.

关于客户

TINE SA is Norway's largest producer, distributor, and exporter of dairy products with 11,400 members (owners) and 9,000 cooperative farms. TINE’s mission is to provide consumers with food that provides a healthier and more positive food experience. TINE has a long history of developing decision-making tools for farmers and collecting data from different dairy products at the farm level. The Norwegian farmers who work with TINE are technologically savvy and have understood the value of collecting data on their animals and dairy production for decades. TINE has conducted structured research to examine the ‘optimal cow’ for dairy production using collected data. They found that focusing on the individuality of a cow enables farmers to discern when each cow is happy, stressed, anxious, and healthy. When their animals are healthier and happier, the quality of their milk is improved, which allows TINE to make even better dairy products while improving the welfare of the animal.

解决方案

To identify the technologies and platforms that would help improve TINE’s insights, predictions, and analyses, TINE brought in the experts at Crayon, an AWS Partner Network (APN) Advanced Consulting Partner and AWS Machine Learning (ML) Competency Partner. Inmeta, a subsidiary of Crayon, worked directly with the customer, providing ideas for data-driven innovation. Crayon approached the innovation project with TINE in four distinct phases. First, the team focused on data readiness, availability, quality, and relevance. During this phase, Crayon concluded that TINE had useful and relevant data but lacked historical timeseries. Next, the team moved on to Methodology and model selection. Crayon chose a simple model initially to demonstrate the use of a convolutional neural network to predict milk production based on the condition on the farm. During this phase, the teams concluded that the model and data were viable for milk production predictions. Then Crayon moved on to revising and refining the modelling process. Based on the finding from the previous phase, the team chose a decision tree model and then expanded to predict the conditions on the farm in the future, which would support better and more accurate forecasting. The model, which uses an ML solution that runs on AWS, predicts cow births and the total number of cows in the herd in addition to milk production. Finally, Crayon established a data lake optimized for analysis and ML models for prediction. This enables further improvements in the ML models while supporting ML initiatives for other applications within TINE.

运营影响

TINE has begun to introduce IoT devices, such as sensors on cows, onto farms to further increase the data scope and better optimize its production processes.
The model provides an automated and accurate prediction of milk production and allows for granularization of the predictions to enable improved production planning for dairy farms and their logistics.
The learnings from the model will also assist TINE and its farmers as they optimize milk production input and management, such as food concentrates.

数量效益