xtractis® Applications

The xtractis® modelling technology extracts the knowledge implicitly embedded in any database. It is therefore universal and can be applied to any problem for which an input/output database of a process is available. No specific knowledge concerning the problem is needed to be able to generate efficient and robust models.

Thanks to its high performance, the xtractis® approach has been successfully used on numerous sensory engineering problems, subjective evaluation problems, descriptive analysis, industrial diagnosis, ...

Segmentation of consumer population

• Identification of groups of consumers with similar liking over a given range of products
• Characterization of consumer segments according to socio-demographic attributes

Optimization of marketing mix scenarios

• Modelling of consumer response to marketing mix scenarios
• Optimization of marketing mix factors to maximize consumer response

Maximisation of liking

• For each segment of consumers, modelling of liking according to sensory attributes of products, instrumental measurements, formulation, or any combination of these data types.
• Identification of drivers
• Identification of product characteristics (sensory profile, instrumental measurements) maximizing the liking of each segment

Formulation optimization

• Modelling of the sensory profile, instrumental measurements and/or production constraints (cost, health concerns) according to formulation.
• Identification of product formulation matching a target sensory profile or set of instrumental measurements while respecting production constraints.

Modelling of tacit knowledge

• Extraction of linguistic rules explaining the decision making processes embedded in a database
• Identification of key factors of decision

Assistance for critical decision making

• Easy validation by a non statician expert, thanks to a report explaining how a prediction has been computed

Design

• Optimization of parametric product definition
• Dimensioning of structures

Process

• Optimization of process parameters
• Diagnosis
• Automatic classification, Clustering
• Detection of flaws, Quality indicator

Virtual Sensor

• Modelling of a costly instrumental measurement according to other cheap instrumental measurements

• Modelling of Structure/Activity Relationships (SAR)
• Diagnosis assistance
• Optimization of drug formulation
• Prediction of toxicity / ecotoxicity

• Prediction of urban density
• Optimal geographical location of resources
• Non linear kriging

Modelling of fraudster behaviour

• Insurance, credit-card, web, welfare benefits
• Detection of fraudster's actions
• Criminal profiling

Natural Sciences

• Climatology
• Agronomy, Agrology
• Oceanography

Medical Sciences

• Epidemiology

Human & Social Sciences

Modelling of consumer behaviour

• Churn score (customer's propensity to leave)
• Non payment score
• Insurance fees according to the predicted damage score, pay-per-drive

Customer's profile optimization

• Discovery of the profile minimizing the churn / non-paiement / damage risks

• Sales forecasts
• Stock optimization