Skip to content

Testing and validation of the performance of predictive models based on nondestructive spectral sensor data (NIR and Hyperspectral data).

Contact Service Provider

Summary

The service evaluates the accuracy and precision of multivariate predictive models, based on nondestructive spectral sensor data (NIR and Hyperspectral data), offering advice for their optimization by chemometric and statistical studies. The service will be supported by the software facilities available, with a multitude of chemometrics dedicated software. The service also includes the possibility of database transfer for their utilization in other instruments or in a sensor network. The service offers the use of reference materials and a reference agri-food sample bank for the evaluation of predictions.

S00271
Version: 1.8

  • horticulture
  • tree crops
  • viticulture
  • food processing
  • collection of test data
  • performance evaluation
  • data analysis
  • conformity assessment
  • AI model training
  • people training
  • software or AI model
  • other (including nothing)
  • Location: Remote
  • Offered by: UCO

E-Mail

Description

The use of NIRS models to predict the quality and safety of agrifood products is a clean, safe, efficient and innovative digital analytical tool. However, production processes are environments in continuous change, from supplier modifications, changes in process conditions, changes in the final product itself, either due to legislative modifications or to meet markets with different needs or requirements, and even changes in instrumentation when acquiring new, more precise and modern devices. Thus, NIRS models are dynamic and need to be updated throughout the life cycle of the instrumentation.This service aims to offer validation and advice on the predictive models based in spectral sensors data that are in use by the company, evaluating their predictive ability through statistic and chemometric studies and advising improvement strategies such as the detection of potentially interesting samples for enlarging databases through the study of their spectral and reference information. We have available a multitude of commercial software through which this assessment can be carried out adjusted to every specific needs. It also aims to assess the potential need for instrumental cloning to be able to use existing databases, which are of high economic value, in newly acquired sensors, through the design of cloning protocols specifically adapted to the user’s needs. The service also includes the use of a large bank of agri-food referenced samples through which the model evaluation and instrumental cloning can be performed.