Machine learning has a special chemistry

Machine Learning has a Special Chemistry
 

“[Creating] the conditions for sustainable agriculture and rural development …will involve …the development of appropriate and new technologies.”

Article 14.2, Agenda 21, United Nations Earth Summit, 1992.

 

Efficient farming practices demand that farmers routinely test their soil for nitrogen and carbon content before making decisions about fertilising their land.

A University of Waikato Computer Science team led by the Dean of Computing and Mathematical Sciences, Professor Geoff Holmes, has developed computerised "machine learning" techniques that can be used to speed up the analysis of data by testing laboratories to help meet these demands. Software exploiting near infra-red(NIR) spectroscopy analysis has reduced the time taken for soil testing from several days to minutes. NIR provides a "blueprint" of a soil sample that can be used to identify the quantity of nitrogen or carbon it contains. Samples can be processed at great speed without the need for "oldfashioned" chemical testing methods, and farmers get fast and cost-effective answers.

This software has now been in routine use at Hill Laboratories in Hamilton for three years. A joint venture between WaikatoLink and Hill Laboratories, Khipu Systems Ltd, has been established to commercialise the software, and a licence sold to another analytical laboratory in the Netherlands. With the expectation that these same techniques can be used to help speed up a range of tests for regulatory compliance, quality control, process control and traceability, Khipu Systems has attracted investment of $1 million to help grow the company.

Meanwhile, the original research team at Waikato has secured further funding to develop machine learning software based on a different technique, gas chromatography, to speed up food testing for pesticide residues and environmental testing for petrol residue in water and soil.

External funding gratefully acknowledged: Foundation for Research, Science and Technology.

MACHINE LEARNING GROUP
DEPARTMENT OF COMPUTER SCIENCE
FACULTY OF COMPUTING AND MATHEMATICAL SCIENCES

This page has been reformatted for printing.