Let your data speak!

Exploring data with Data Mining methods

You have huge amounts of data from different sources: instrumental (near infrared spectroscopy (NIR), mid infrared (MIR), Raman, Hyperspectral imaging, HPLC chromatography, GC, etc.), process parameters, physico-chemical measurements, sensory profiles… with a large number of samples and / or variables. And understanding and extracting relevant information from them seems very complex to you …

There are exploratory methods that will allow you to find meaning in this data, to detect atypical samples, outliers and to identify groups of individuals or strong trends.

You want to

Visualize

atypical samples, clusters, tendencies…

Understand

leverages of quality of a process, complementarity of measurements

Interprate

spectral mixing, correlation between blocks of data

data mining - information extraction

Among them Principal Component Analysis (PCA) is the most classic multivariate method of Data Mining. These exploratory models constitute a precious help to define and optimize leverages of quality in various applications.

Other more specific methods of signal deconvolution can also be applied for your spectroscopic data, particularly in the fields of chemistry or pharmaceutical industries. Among these methods, we will find Multiple Curve Resolution (MCR) or Independent Component Analysis (ICA). They allow a better interpretability of model components by focusing on the extraction of pure spectra.

If you have multiple blocks of data from different analytical techniques and/or different sensors, the complexity is even greater. It is therefore interesting to combine these data blocks to extract even more information, in particular the joint information between all blocks and specific information of each block. The exploratory analysis is then carried out using multi-blocks methods.

Thanks to our knowledge in Chemometrics and Machine Learning on data from R&D, pilot tests or measurements on industrial process lines, we assist you to explore your data and helps you “make it speak”.

data mining and information extraction
  • Multivariate exploratory analysis (Principal Component Analysis –PCA-)
  • Signal deconvolution (Independent Component Analysis (ICA), Multivariate Curve Resolution (MCR-ALS), SIMPLISMA, etc.)
  • Clustering of unknown groups (Unsupervised classification methods such as HCA, kNN, etc.)
  • Data fusion thanks to multi-block analyzes (ACCPS, ACOM, etc.)
  • Multivariate variance analysis (A-PCA, ASCA)

Software implementation

We develop your calibrations with all the Chemometrics software on the market: Unscrambler®, SOLO® or PLS_Toolbox® in the MATLAB® environment, SIMCA®, etc… but also free software (R, Python…).

Our expertise for the analysis of your data

With over 15 years of experience in data analysis (Chemometrics and Machine Learning) from a wide variety of sources, the experts in our teams support you at every stage of your projects.

They talk about Ondalys


« Ondalys brought us skills in terms of multi-block data processing. »

Dr Jean-Michel ROGER, Researcher at INRAe - Irstea

Register to
receive Ondalys newsletter

English